《ChiefMartec&MartechTribe：2025年营销技术展望报告（英文版）（111页）.pdf》由会员分享，可在线阅读，更多相关《ChiefMartec&MartechTribe：2025年营销技术展望报告（英文版）（111页）.pdf（111页珍藏版）》请在三个皮匠报告上搜索。

1、Martech for 2025by Scott Brinker and Frans RiemersmaSponsored by:December 3,2024&Martech for 2025Scott Brinker is the editor of the chiefmartec blog,covering the intersection of marketing and technology for over 15 years.He serves as the vice president of platform ecosystem at HubSpot.He is also the

2、 author of the best-selling book Hacking Marketing and co-author of The New Automation Mindset.He jointly produces the MartechMap marketing technology landscape with Frans.Frans Riemersma founded MartechTribe,a company specializing in Martech research and benchmarking.With 30+years in consultancy,he

3、 combines qualitative expertise with quantitative Martech data on stacks,vendors,and requirements in a proprietary Martech Data Warehouse.He is the author of A Small Book on Customer Technology and co-author of Marketing Tech Monitor,Customer Technology Sector Trends,and Hello Firstname.In collabora

4、tion with Scott,he also co-produces the MartechMap,an overview of the marketing technology landscape.This report is sponsored by GrowthLoop,Hightouch,MetaRouter,Progress,and SAS.AuthorsSponsors 2024 Marketing Technology Media,LLC,MartechTribe B.V.,and MartechUnited B.V.Graphic design,information des

5、ign,and illustrations by Angela Ribeiro da Silva(Dattura Studio)&Martech for 20251.How AI is Reshaping the Marketing and Martech Environment GenAI Boosts Martech Growth in 5 SegmentsThe First 3 Segments:Indie Tools,Incumbent and Challenger PlatformsThe 4th Segment:Instant Software The 5th Segment:Se

6、rvice-as-a-Software 1629710 17252.Foundations for an AI Strategy The Evolving Universal Data LayerThe(Underdeveloped)Universal Content LayerAPI Composability as AI Agent Building Blocks IntroductionFive Perspectives on Martech for 2025GrowthLoop:End-to-End Marketing on Your Data Cloud with AIHightou

7、ch:Why the Next Wave Beyond CDPs is AI DecisioningMetarouter:A CDPs Best Friend:“Shifting Left”Data QualityProgress:Lessons of Composability for Marketing OperationsSAS:Filling the Gaps in Governance for Generative AI3.How Marketers Are Using GenAI Today Popular Martech GenAI Use CasesNew AI Tools v

8、s.AI Embedded in Current Martech ToolsFrequency of GenAI Tool Usage by Use CaseGenerative AI Policy and Impact on UsageDemographics of Survey Respondents2933364344495459626464748693101Content&1Martech for 2025We are absolutely at a place where,if AI development completely stopped,we would still have

9、 5-10 years of rapid change absorbing the capabilities of current models and integrating them into organizations and social systems.I dont think development is going to stop,though.Ethan Mollick,The Wharton SchoolIs AI overhyped?Yes.Is AI a massive disruptor?Also yes.The logic to that seeming parado

10、x is Amaras Law:we tend to overestimate the effect of a technology in the short run and underestimate its effect in the long run.P and Webvan,epically failed dot-com businesses from the late 1990s,overestimated e-commerce in the short run.At the peak of the dot-com boom in 1999,worldwide e-commerce

11、transactions totaled$100 billion,a mere 0.3%of the worlds GDP.But by the end of 2018,global ecommerce totaled$25.6 trillion1 B2B and B2C combined a whopping 29.7%of the$86.2 trillion2 global GDP.Amazon went from being a$5 billion company in 2001,its future deeply discounted by Wall Street,to now bei

12、ng the 5th largest company in the world with a market cap of$1.8 trillion.How many of you,back in 1999 expected Amazon to reach that scale?Did you invest?Did you hold on to your investment through the dot-com crash?1 https:/unctad.org/fr/isar/news/global-e-commerce-hits-256-trillion-latest-unctad-es

13、timates 2 https:/ for 2025The thing to remember about the Hype Cycle,of course,is that its a measure of hype not a measure of the underlying technology.While the hype bobs wildly up and down,the technology behind it steadily marches forward.And if so,are you reading this report from your boat with a

14、 shoddy connection in the tropics,three months after we published it?3 Thats Amaras Law.Amaras Law has repeated itself time and again throughout the history of technology.Grand expectations when a new innovation first appears,quickly hyped to an extreme.Disappointment when those expectations fail to

15、 materialize quickly.But steadily the technology grows,improves,and is adopted in more and more applications.Until one day,years later,its scale and impact far exceeds the grandest expectations that we hoped for at its beginning.If you were to visualize Amaras Law on a graph,it would look a little l

16、ike or a lot alike Gartners famous Hype Cycle:3 A nod to JP Castlins excellent Strategy in Praxis newsletter.The Hype CycleGartners famous(infamous?)Hype Cycle is a visual approximation of Amaras LawTechnology TriggerTrough of DisillusionmentPeak of Inflated ExpectationsSlope of EnlightenmentPlateau

17、 of ProductivitySource:Gartner/chiefmartec&3Martech for 2025The only difference with Amaras Law is that“Plateau of Productivity”isnt the summit.Its a stop along the way,followed by another climb upwards often with its own squiggly hype curve followed by another,and another,and so on.That ascent over

18、 decades took us from the 3-ton mainframes of the 1960s to having 10 million times more computing power in the palm of your hand with todays mobile phones.As youve no doubt surmised,this is the journey were on with AI.But there are a couple of key differences this time.First,the rate of change is mu

19、ch,much faster.In previous reports,we described this as the compression of the Hype Cycle.A pattern that used to take 5-10 years to play out,from technology trigger to productivity plateau,now happens in half that time.ChatGPT is barely 2 years old,yet it already has over 200 million users per week.

20、Second,AI is not really a single Hype Cycle,but a multitude of them,all at different stages,many entangled with each other.“Generative AI”is a massively large field with different use cases all at different stages of maturity.All of them inherit advances in the rapidly improving The Hype Cycle and T

21、echnology DevelopmentWhile the“hype”has its ups and downs,the underlying technology continues to advanceTechnology TriggerTrough of DisillusionmentPeak of Inflated ExpectationsSlope of EnlightenmentPlateau of ProductivityTechnologySource:chiefmartec&4Martech for 2025foundation models from OpenAI,Ant

22、hropic,Google,Meta,etc.Yet they each have their own technology innovations wrapped around them.And generative AI itself is just one part of a larger AI universe,where the myriad of machine learning(ML)techniques and applications have been advancing for years.Useful applications of ML are already emb

23、edded in almost every software product you use today.With greater scale of data and computing power,the capabilities of ML continue to grow at the exponential rate of Moores Law.Prompt to ChatGTP:“Generate a tree diagram for me of the many different ways in which generative AI is being applied.”Appl

24、ications of Generative AIVirtual Reality ContentTranslatorsContent Creation3D ModelGenerationPhotographyEnhancementCreative WritingFilm andSpecial EffectsSynthetic Data for TrainingData AugmentationAI-Generated Test CasesCode TranslationDeepfakesText GenerationVideoGenerationGenerativeAIImageGenerat

25、ionGenerativeCodeChatbots and Conversational AgentsFashion and StyleCode CompletionSound EffectsCreationSpeech-to-Speech TranslationSummarizationAudioGenerationVoiceSynthesisAnimatedCharactersArt andDesignMedical ImagingMusic CompositionAutomatedProgrammingSource:chiefmartec&5Martech for 2025All of

26、this combines to make it very difficult to say,“AI is overhyped.”Some of it is,for sure.But much of it is already being deployed throughout marketing,with new functionality and use cases continually being added to the repertoire of modern marketing.Quoting Ethan Mollick,one of the leading observers

27、of AIs ever-morphing abilities,even“if AI development completely stopped,we would still have 5-10 years of rapid change absorbing the capabilities of current models and integrating them into organizations and social systems.”But its not going to stop.Welcome to Martech for 2025.&6Martech for 2025Its

28、 like riding a roller coaster,isnt it?Being a marketer in this AI era brings a flood of mixed emotions.The thrill of an exhilarating new ride.The nervous excitement from not knowing whats around the next curve.The anxiety as you consider the possibility that the whole thing might just fly off the ra

29、ils and possibly take your career with it.Yet youve already purchased your ticket,climbed on the trolley,and are being dragged up the lift hill.Too late to turn back now.AI is reshaping marketing and martech.And while were not prone to hyperbole,we do believe there will be significant real-world cha

30、nges that marketers and marketing operations leaders will have to face with this technology in 2025.So while consciously avoiding a bunch of bull-slinging,heres the martech environment for AI that we see now and expect in the year ahead.1.How AI is Reshaping the Marketing and Martech Environment&7Ma

31、rtech for 2025In our State of Martech 2024 report in May of this year,we revealed our(in)famous marketing technology landscape had grown a whopping 27.8%from 2023 to 2024.The total number of martech tools we tracked grew from 11,038 to 14,108 in just one year.The vast majority of this growth came fr

32、om an explosion of AI-powered specialist tools in the“long tail”of the market:Number of Martech Software Apps Since 2011GenAI Boosts Martech Growth in 5 Segments0200040006000800010000120001400016000201120122013201420152016201720182019 2020 2021 2022 2023 202415035094718763874538168297040800099321103

33、8141069.304%growth over 13 years 27.8%growth in the last 12 months 41.8%CAGR Source:chiefmartec&MartechTribe&8Martech for 2025New AI Products in 2024Top 14 New GenAI Based Tools in MartechTailTorsoHeadScale(revenue&install base)10s100s1000sMartech productsNew AI-Native ProductsAI into Existing Produ

34、cts+3,068 new products(77%AI-native)ManagementAdvertising&PromotionsData Commerce&SalesSocial&RelationshipsContent&ExperienceContent MarketingSales AutomationEnablement&IntelligenceVideo MarketingBusiness/CustomerIntelligence&DataScienceLive Chat&ChatbotsSocial Media Marketing&MonitoringCollaboratio

35、nCMS&Web ExperienceManagementAudience/Marketing Data&Data EnhancementCustomer ExperienceService&SuccessEmail MarketingSEOMobile AppsOptimizationPersonalization&Testing33.1%11.3%6.3%5.4%5.2%3.5%3.5%3.4%3.3%3%2.5%2.3%2.1%1.6%Sales use casesClient meetings,notesLead sourcing&OutreachLead ScoringPitch D

36、ecksGenAI Data use casesData Infrastructure,Storage&UnificationData Interpretation&Chat with DataData Sourcing&ExtractionSource:chiefmartec&MartechTribeSource:MartechTribe&9Martech for 2025We tracked 2,324 new AI-native products,most of them in the content marketing and sales automation categories.A

37、nd that didnt include hundreds of custom GPTs in OpenAIs GPT Store focused on marketing use cases.Consolidation continued to happen elsewhere in the market.But the number of new martech solutions being tracked far exceeded the number that went away.The martech landscape continues to be a seeming par

38、adox of simultaneous consolidation on one end of the spectrum and new growth on the other.As F.Scott Fitzgerald famously wrote,“The test of a first-rate intelligence is the ability to hold two opposed ideas in the mind at the same time,and still retain the ability to function.”Thats been the martech

39、 motto.But will AI ultimately break that pattern?&10Martech for 2025Of course,its not just new martech products providing AI functionality to marketers.The past year has seen a frantic rush by existing martech companies to add new AI capabilities to their products.Adobe,HubSpot,Microsoft,Salesforce,

40、SAS,etc.all added a slew of new features with generative AI as well as more features leveraging good,old-fashioned analytical AI with machine learning.Through one lens,this sets up a repeat of the age-old struggle of startups vs.incumbents.The“head”of the martech long tail the dominant martech platf

41、orms of the past decade and the thinly stretched“tail”of new martech startups are competing as they always have.AI just happens to be the field of battle today.But we think theres an important nuance here.Indie Tools,Challenger and Incumbent PlatformsThe First 3 Segments:Indie Tools,Incumbent and Ch

42、allenger PlatformsIncumbent PlatformsChallenger PlatformsIndie ToolsSubsumingDisruptingM&ASource:chiefmartec&MartechTribe&11Martech for 2025Only a relatively small percentage of these AI-native startups are intentionally trying to compete with the incumbents.Most are just experimenting with what gen

43、erative AI engines from OpenAI,Google,Anthropic,Meta,and others now make possible.Theyre building small,stand-alone tools that complement existing platforms by automating or augmenting tasks that marketers have done manually around those platforms.Examples include:Headlime AI-powered marketing copyw

44、riting with templates browse.ai AI-powered web site scraper for data and monitoring tl;dv AI-powered notetaker for Zoom,Google Meet,MS Teams Vizly AI-powered data analysis and visualization Howler AI AI-powered personalized intro lines for sales reps SpeakAI.co AI-powered transcription and translati

45、onWe call these indie tools because many of them have little or no institutional VC funding.(Its remarkable how much can be built so quickly with so little investment now a point well come back to in a little bit.)Not only are most indie tools not trying to compete with the large incumbent platforms

46、.Many are actually seeking to leverage and integrate with them in their ecosystems.Hugh Durkin,the founder of AppM,a service that tracks app ecosystems across B2B platforms,has reported an explosion of“AI”tools added to these marketplaces this year.Non-AI vs.AI Apps0%100%Added inOct.2024Overall in20

47、2471.4%90.1%28.6%9.9%AI AppsNon-AI AppsSource:Hugh Durkin,AppM,October 2024&12Martech for 2025To help visualize the incredible scale of this AI-native indie tool space,we produced a special genAI version of our Marketing Technology Landscape earlier this year that highlights the collection of these

48、indie tools that we discovered over the past year.You can browse an interactive PDF version here.However,some of these AI startups are intentionally seeking to disrupt the major martech incumbents.Andreesen Horowitz published a barnburner of a blog post in July called“Death of a Salesforce”:Why AI W

49、ill Transform the Next Generation of Sales Tech that speaks to the strategy of these challenger platforms:We believe AI will so fundamentally reimagine the core system of record and the sales workflows that no incumbent is safe.Instead of a text-based database,the core of the next sales platform wil

50、l be multi-modal(text,image,voice,video),containing every customer insight from across the company.An AI-native platform will be able to extract 2024 MarTech GenAI LandscapeSource:2024 MarTech GenAI Landscape,chiefmartec&MartechTribe&13Martech for 2025more insight from a customer and their mindset t

51、han we could ever piece together with the tools we have today.Sales workflows will fundamentally change.With AI,sales teams will no longer need to spend endless hours researching new leads or prepping for calls AI will be able to do it in seconds.Reps wont have to suss out the readiness of potential

52、 customers because AI will have automatically compiled a ranked list of primed buyers,and will keep it constantly updated.Need personalized marketing collateral for a deal?Your AI wingman will produce whatever assets you need and feed you live tips while youre on a call to help you close.The example

53、s they cite are Clay for prospecting data intelligence,11x for autonomous AI agents serving as inbound and outbound SDRs,Naro for sales enablement,Day as an AI-native CRM,and People.ai for a new generation of salesforce automation(“salesforce”as the employees in a company responsible for selling,not

54、 the blue cloud SaaS pioneer).The advantage these AI-native challenger platforms have is that they can rethink everything data models,workflows,user interface,pricing and package,etc.for a new AI-powered world.They are not constrained by technical backward compatibility or the dynamics of Clayton Ch

55、ristensens Innovators Dilemma biasing their business model choices.However,these challengers have their own disadvantages to overcome.Most are incomplete and require integrations with multiple other products to cover the capabilities customers require.Because they have small user bases,they often ha

56、ve access to small amounts of data to train their AI features.In contrast,the incumbents theyre targeting have large user bases and are deeply integrated in those customers tech stacks.Displacing these incumbents la rip-and-replace is a daunting proposition,especially given that these challenger pla

57、tforms carry existential risk:90%of startups fail.44 https:/ for 2025Where challenger platforms gain a tailwind is when the fundamental nature of the category may change as a result of AI a true paradigm shift.For instance,competing with existing DXPs such as Adobe,Optimizely,and Sitecore is a hard

58、uphill climb unless websites themselves shift to become something very different.If avatar agents or dynamic content canvases engage customers outside of traditional web pages and site navigation,the very nature of a DXP gets disrupted.That opens the door for a new generation of martech leaders.The

59、role that AI agents could play in disrupting how work gets done and customer experiences get delivered will likely open many such doors.But the incumbents are not standing still.Salesforces bold moves with Agentforce and Generative Canvas.HubSpots Breeze co-pilots and agents,and co-founder Dharmesh

60、Shahs agent.ai(a“professional network for AI agents”).Microsofts embedding of Copilot throughout their Dynamics product suite.Adobes GenStudio is a seamless AI-driven content supply chain.Indie Tools,Challenger and Incumbent PlatformsIncumbent PlatformsChallenger PlatformsIndie ToolsSubsumingDisrupt

61、ingDisadvantagesLimited resources and scale,low defensibilityHard to dislodge incumbents,viability risk,limited dataConstrained by backwards compatibility and existing revenue structuresThousands of stand-alone AI specialist tools that do one thingNew AI-native platforms that aspire to disrupt exist

62、ing platform incumbentsDominant martech platforms rapidly embedding AI features in their products,organically and via acquisitionsAdvantagesCreatively use AI to do one thing really,really well with specializationDesigned from the ground up with new AI paradigms,free from innovators dilemmaLarge user

63、 bases,deeply embedded in tech stacks,big cash flows/reserves,huge amounts of data for trainingM&ASource:chiefmartec&MartechTribe&15Martech for 2025As Sonya Huang and Pat Grady of Sequoia Capital recently wrote in an article,Generative AIs Act o1:The Agentic Reasoning Era Begins,“The classic battle

64、between startups and incumbents is a horse race between startups building distribution and incumbents building product.Can the young companies with cool products get to a bunch of customers before the incumbents who own the customers come up with cool products?”The incumbents are moving quickly to b

65、uild cool AI capabilities.They are also leveraging their equity and capital reserves to acquire hot indie tools and rising challenger platforms to make non-linear jumps forward.And theyre already leveraging their large customer bases their distribution to roll out that new AI functionality at scale

66、to shut out the challengers before they get their foot through the door.One could easily assume that the incumbent platforms will come out on top.Yet the incumbents have hurdles to overcome that arent about the technology per se.Clay Christensens Innovators Dilemma will be a real struggle for them t

67、o balance their commitments and expectations with existing customers and the reliable financial engine of their existing pricing and packaging with new product paradigms and new pricing models.(Dharmesh Shah,the co-founder and CTO of HubSpot,has shown one way of escaping that dilemma by producing pr

68、oducts such as ChatSpot and,more recently,Agent.ai adjacent to HubSpots core platform.)Shifts in Pricing ModelsmodelSaaSIaaS/PaaSservice-as-a-softwareproduct formapplicationsservicesAI agentsaligned toemployeescompute/data resourcesgoalscostsfixed,based on number of usersvariable,based on resource u

69、tilizationvariable,based on outcome demandcost-related riskunderutilizationpoor optimization or forecastingmiscalculated value or demandOutcomesUsageSeatsSource:chiefmartec&16Martech for 2025But the bigger hurdle for them may simply be mindset and their ability or inability to truly imagine how this

70、 new world will differ from the old one.To quote a bit more from Huang and Grady:Twenty years ago the on-prem software companies scoffed at the idea of SaaS.“Whats the big deal?We can run our own servers and deliver this stuff over the internet too!”Sure,conceptually it was simple.But what followed

71、was a wholesale reinvention of the business.EPD went from waterfalls and PRDs to agile development and AB testing.GTM went from top-down enterprise sales and steak dinners to bottoms-up PLG and product analytics.Business models went from high ASPs and maintenance streams to high NDRs and usage-based

72、 pricing.Very few on-prem companies made the transition.5What if AI is an analogous shift?Could the opportunity for AI be both selling work and replacing software?So who will win?Its not clear.And practically speaking,its not a strictly zero-sum game in the near-term.For 2025,we believe most markete

73、rs will benefit from all three:Leverage the new AI features being released by your existing incumbent platform(s).Adopt indie tools that complement those platforms and let you experiment with new AI capabilities quickly and inexpensively.And keep an open mind to creative new approaches championed by

74、 AI challenger platforms.Whether they ultimately succeed as companies or not,they are likely augurs of the future.5 Holy acronym city,Batman!EPD=enterprise product development.PRD=product requirements document.GTM=go-to-market.PLG=product-led growth.ASP=average selling price.NDR=net dollar retention

75、.AB=version A and version B of something,being tested against each other.&17Martech for 2025But true disruption often comes at you from an angle you arent expecting.While commercial incumbent and challenger platforms are battling it out,we believe one of AIs biggest impacts will be the explosion of

76、custom“software”in businesses of all sizes not just enterprises.We put“software”in scare quotes because,increasingly,people will create software programs through AI without even realizing it.(Should we call it“instant software”in a nod to the magic of“instant coffee”just add water,or in this case,ju

77、st add intent?)But first,lets acknowledge that even conscious custom software development is on the rise.While public LLM services have become extremely popular ChatGPT,Claude,Gemini,etc.they are generic in their reasoning and training data.Custom AppsThe 4th Segment:Instant SoftwareIncumbent Platfo

78、rmsChallenger PlatformsIndie ToolsSubsumingDisruptingM&ACustom AppsReplacingAugmenting&ReplacingSource:chiefmartec&MartechTribe&18Martech for 2025To harness their power in ones own digital operations and customer experiences,businesses need to tailor them to their proprietary data and context-specif

79、ic logic.Tailoring can be done through a variety of methods:training ones own model,fine tuning an existing model,and/or using RAG(retrieval-augmented generation).RAG,which is the most common method today,looks up data from internal databases and feeds it behind-the-scenes into the prompts given to

80、the LLM engine.The LLM generates its response using that retrieved data as input,augmenting its own generic knowledge.Hence,retrieval-augmented generation.RAG implementations may also check or manipulate the output from the LLM to provide further guardrails on the end result.All of this is done by b

81、uilding custom software to connect these pieces together,often using Python,which has become the preferred programming language for AI.These custom generative AI implementations can be run stand-alone or incorporated into larger software apps.We often see teams follow a hack-pack-stack progression h

82、ere.Quickly test an idea in a stand-alone fashion(hack).If proven feasible,build it the“right”way and deploy it into production(pack).If it delivers sustained value,optimize it for long-term maintainability(stack).The Hack-Pack-Stack ProgressionTest an idea quickly to determine if it is technically

83、feasible and if it will deliver the intended value/outcome;an experimentOnce the idea is validated if successfully addresses a clear need implement it the proper way and deploy it in full productionRefactor into a scalable,low-maintenance version and integrate it into the ecosystem of your tech stac

84、k(your digital ops”DNA”)best-of-needbest-of-speedbest-of-breedPackHackStackSource:chiefmartec&MartechTribe&19Martech for 2025AI coding assistants,such as GitHub Copilot,Gemini Code Assist,Amazon Q Developer,Cursor,and more,are accelerating the speed and quality of proper software development.Combine

85、d with the rise of composability with APIs and organization-wide accessibility of data through cloud data warehouses,companies are more willing and able to build pieces of custom software as part of their overall digital operations.As an extreme example,the fintech giant Klarna announced earlier thi

86、s year that they were ditching Salesforce and Workday to build their own custom CRM and HCM applications using AI and composable cloud services.6 While theres been skepticism about the wisdom of such a“build it all yourself”strategy,the fact that such a move is even conceivable is a testament to bot

87、h the improved economics of custom development and the perceived business advantage of more tailored digital operations in the AI era.6 https:/ for 2025What should you build vs.buy?Largely its a question of comparative advantage.Where commercial companies in the market have greater experience and ex

88、pertise in building a particular kind of product or platform,its often most economically rational in both direct costs and opportunity costs to buy.Where can your company generate unique value from its experience and expertise expressed in software?Those may be a candidate for building.Its where the

89、 expertise in your company for those specialized needs is greater than the more generalized expertise in the market at large with commercially packaged products.As a new technology winds its way up and down the Hype Cycle as a number of AI capabilities currently are the rationale for building vs.buy

90、ing often evolves.In the early stages of the technology,we buy to learn from those companies who have specialized in the technology.If we build,its often to prototype exploratory ideas.The Technology Hype Cycle Buy to learn Build to prototypeExpertiseMarket CompanyExpertiseCompany Market Buy to comp

91、ose Build to integrateSource:MartechTribe&21Martech for 2025In later stages,we buy to acquire viable and proven capabilities that we can then compose into our business operations and customer experiences.We build the unique components that leverage our domain expertise and integrate everything toget

92、her.But the greatest quantity of custom software ahead wont be built by professional developers.It will be apps,automations,and analyses created by business users through“no-code”interfaces and AI.Weve already seen an explosion of such citizen development,as Gartner labeled it,over the past 10 years

93、.Products such as Airtable(database apps),Webflow(web apps),and Workato(workflow automation)have empowered business ops teams including marketing operations and revops to self-serve many of their own custom software needs.Gartner predicted that by 2025,70%of new applications developed by organizatio

94、ns will use low-code or no-code technologies,up from less than 25%in 2020,that are“driving the increase of citizen development,and notably the function of business technologists who report outside of IT departments and create technology or analytics capabilities for internal or external business use

95、.”7And that prediction was made before the genAI revolution broke out.Now,we have a whole new wave of no-code AI agent builders,such as Googles Vertex AI Agent Builder,Relevance AI,and Kore.ai.But even with no-code tools,these are all explicitly created software“programs.”Business users might not th

96、ink of themselves as developers,but they recognize that they are building an app,an automation,or an agent when they use those tools.However,when you ask an AI agent to do something for you and it creates a custom program on-the-fly to execute it completely behind-the-scenes,and by default not visib

97、le to the user who made the request is that custom software?We would argue that it is.7 https:/ for 2025Such automatic or even ambient creation of software programs by AI agents will be faster and cheaper by orders of magnitude than any software development process that came before.And because the i

98、nterface of these AI agents will be accessible to any business user,the number of people triggering the creation of these dynamic programs will asymptotically be everyone in the company and,through customer-facing agents,potentially many more.Beyond the millions of human-built no-code apps,there wil

99、l be billions of AI agent dynamic programs.This explosion of unseen custom“software”has already begun,and we expect it will accelerate in 2025 as AI agents proliferate.The AI agent is,in a very real sense,a software developer.(Albeit one without a love for pizza and Star Wars.)The Democratization of

100、 Marketing Technologytimediscipline expertsdomain expertspower usersusersambient,automaticmarketing operationsmarketersITcodelow-codeno-codeAI agent dynamic programswhohowcost,expertise requiredvolume,production ratedemocratization of marketing technologySource:chiefmartec&23Martech for 2025This wil

101、l change the composition of tech stacks.Not only will companies draw from the long tail of commercial applications,such as those we catalog in the ever-bulging marketing technology landscape.They will complement and augment them and in some cases replace them with a growing quantity of custom“softwa

102、re”that they build for their unique operations and customer experiences.If the commercial martech landscape has been a“long tail”distribution a small number of major martech vendors in the head and torso,followed by a large number of martech startups,specialists,and challengers in the tail we see th

103、is explosion of custom IT-built,citizen-built,and agent-built software as an effectively infinite extension of software choice.We call this a hypertail distribution.With due respect to Klarnas decision to ditch major commercial platforms to roll their own,we actually expect that most tech stacks wil

104、l incorporate both commercial and custom software.Commercial platforms that help aggregate and orchestrate custom software providing cohesion and governance to the plethora of apps,agents,The HypertailTailTorsoHeadScale(revenue&install base)10s100s1000sMartech softwareHypertail1,000,000s1,000,000,00

105、0sAgent-BuiltSoftwareCitizen-BuiltSoftwareIT-BuiltSoftware(.)Source:chiefmartec&24Martech for 2025However,we suspect this may be the turning point where the number of commercial apps in the tech stack peaks and future growth of the stack which overall we think could be exponential will come from cus

106、tom software,a cornucopia of custom apps,agents,and automations.and automations operating in this“big ops”environment will be especially valuable as coordinating centers of gravity.Custom vs.Commercial SoftwareQuantity of Marketing SoftwareMainframe EraPC EraSaaS EraCustom SoftwareCommercialSoftware

107、AI EraTimeSource:chiefmartec&25Martech for 2025The 5th Segment:Service-as-a-SoftwareBut wait,theres more.So much software to date has been tools to help people do their work better,whether through increased efficiency or greater creative range or applying their talents to digital assets,data,experie

108、nces,media,etc.that inherently required a digital interface.Software has been an assist to labor.But with the rise of agentic AI capabilities where a new generation of AI-powered software can actually do more of the work autonomously or with minimal supervision its becoming possible for software to

109、increasingly serve as labor.Services-as-SoftwareIncumbent PlatformsChallenger PlatformsIndie ToolsSubsumingDisruptingM&ACustom AppsReplacingAugmenting&ReplacingServices-as-SoftwareSubstitutingManaging&SubstitutingSource:chiefmartec&MartechTribe&26Martech for 2025That is the gateway to a much,much la

110、rger disruption than challenger platforms vs.incumbent platforms in the software industry.As Grady and Huang from Sequoia Capital noted(emphasis our own):The cloud transition was software-as-a-service.Software companies became cloud service providers.This was a$350B opportunity.Thanks to agentic rea

111、soning,the AI transition is service-as-a-software.Software companies turn labor into software.That means the addressable market is not the software market,but the services market measured in the trillions of dollars.The race is on to reinvent every nook of that multi-trillion dollar services market

112、by providing more services wrapped through AI agents and software interfaces.The clever term for this is service-as-a-software,the new kind of“SaaS.”Existing marketing services will certainly be impacted by this,as you can expect current agencies and a new wave of challenger agencies to leverage AI

113、to change the economics,speed,and scale of the services they have historically provided:market research,audience segmentation,media planning and purchasing,creative production(if not the creative itself),etc.But there will also be new kinds of AI-powered service providers that essentially replace bo

114、th internal software products and the labor required to use them.Instead of in-house tools and talent,there will be options to purchase outsourced“outcomes”from these providers.Why buy a set of content distribution tools,and hire and train a team to use them,if a highly cost-efficient(thanks to AI)s

115、ervice provider can do the work for you quickly,easily,effectively and may only charge you for successful outcomes such as views,engagements,click-throughs,leads,purchases,etc.?For martech and marketing operations,these new kinds of service-as-a-software options will change the make-up of tech stack

116、s and the processes orchestrated around them.&27Martech for 2025As venture capitalist Tomasz Tunguz recently wrote,“When AI products are sold as services,they replace in-house labor.This changes internal processes.When the internal processes change,the opportunity to replace the system of record ari

117、ses because the existing workflows are no longer relevant.”Service-as-a-software options will,in a very real sense,be both complements and competitors to software companies and internally-built custom software.This is not a far-future prediction.Production instances of these software-as-a-service so

118、lutions are already in-market,such as Sierra.ai for“outsourced”customer service interactions(using a pay-per-resolution pricing model).We expect to see many more in the year ahead.The HyperscalersIncumbent PlatformsChallenger PlatformsIndie ToolsSubsumingDisruptingM&ACustom AppsReplacingAugmenting&R

119、eplacingServices-as-SoftwareSubstitutingManaging&SubstitutingHyperscalersPowering the 5 SegmentsExternally sourcedInternally builtSource:chiefmartec&MartechTribe&28Martech for 2025With this full view of the five segments of martech solutions in the AI era,its worth pointing out that all of them are

120、powered by the hyperscalers:Amazon AWS,Google Cloud Platform,and Microsoft Azure.This opens up interesting opportunities to use those underlying infrastructure providers as a more explicit substrate for coordinating across ones internal and external stack.This is essentially the pattern that has eme

121、rged with cloud data warehouses(and cloud data lakes and cloud data lakehouses):a common substrate of data and software APIs that enable marketers to mix and remix insights and functionality across their organization in novel use cases.&29Martech for 2025The Evolving Universal Data Layer“You cant ha

122、ve an AI strategy without a data strategy.”A number of highly respected experts across the industry have expressed that opinion over the past year,and we agree with them wholeheartedly.Your data is what turns generic AI algorithms with both generative AI and classic machine learning(ML)into differen

123、tiated and relevant capabilities that give your business a competitive advantage.From a martech stack perspective,the key to this is a universal data layer that aggregates data from all the different applications in your stack and makes it available for any other app to use.For many companies,this i

124、s being accomplished with a cloud data warehouse(CDW)or cloud data lakehouse(CDL)the latter being able to handle both structured data,as classic data warehouses have,and unstructured data,which has typically been stored in a more open data lake.Hence the“lakehouse”portmanteau.While some marketing or

125、ganizations are using customer data platforms(CDPs)or other“data clouds”as a shared data layer,with more marketing-specific capabilities,more and more of those solutions now actually sit on top of a cloud data warehouse/lakehouse.These are often called“composable CDPs”because they operate on data th

126、at is composed from other data storage sources.In our State of Martech 2024 report,71%of the martech and marketing ops professionals who took our survey on composability reported that they have a cloud data warehouse/data lake in their martech stack,such as Snowflake,Databricks,Google BigQuery,Amazo

127、n Redshift,etc.2.Foundations for an AI Strategy&30Martech for 2025Out of those,61.3%said that more than 50%of the apps in their martech stack were integrated with it.The advantage of a non-marketing-specific cloud data warehouse/lakehouse layer is that it can span the entire organization.This is a h

128、uge unlock for marketing,as it gives marketers access to data from customer touch points that are managed by other departments:sales,customer service,finance,digital product operations,etc.These broader datasets provide richer customer insights and can be used to power better marketing campaigns and

129、 customer experiences.These cloud data warehouses/lakes/lakehouses are the backbone of what has become known as the“modern data stack”for organizations.The stack also includes tools for ingesting data,transforming data,cleaning data,enriching data,governing data,and feeding data into tools for data

130、science,business intelligence,machine learning,and front-line business applications.Do you use a cloud data warehouse/data lake with your martech stack?15.5%8.3%47%29.2%Yes,we send data one-way,from martech apps into our warehouseYes,we send data one-way,from our warehouse into martech appsYes,we se

131、nd data both ways,between apps and our warehouseNo,we dont use a data warehouse with any of our martech appsSource:2024 Martech Composability Survey,chiefmartec&MartechTribe&31Martech for 2025That last piece feeding data from the modern data stack into applications such as CRMs,marketing automation

132、platforms(MAPs),digital experience platforms(DXPs),etc.that directly engage or support customer interactions makes this more of a circular data ecosystem.Those apps are both generating data that is pushed into the universal data layer as well as pulling other aggregated and processed data out from i

133、t.The Circular Data EcosystemModern Data StacksApps&OpsData Security&GovernanceData Workflow&AutomationData Observability(Data Quality)Metadata Management(Data Catalog)Data Transformation&ModelingData LakehouseData LakeData WarehouseData in*structured data*unstructured dataModern Data StackData Ops

134、EcosystemConnectors(iPaaS)Event StreamingExternal DataBusiness IntelligenceReverse ETLGenerative AIMachine LearningData SciencePrep&QueryETL(ELT)OrchestrationData OutStoragelocation formatacessdata flowuse casedata usageinterfacechallengeDataPastPresenton premstructuredfew users1-way(linear)analytic

135、slow utilizationcode(SQL)big datacloudstructured&unstructuredmany users&apps2-way(circular)analytics&operationshigh utilizationUI&AI for business usersbig opsApps&OpsData Security&GovernanceData Workflow&AutomationData Observability(Data Quality)Metadata Management(Data Catalog)Data Transformation&M

136、odelingData LakehouseData LakeData WarehouseData inSource:chiefmartec Source:chiefmartec&32Martech for 2025This evolution of the data layer has catalyzed many changes.By moving data into the cloud,its become easier to integrate with all our cloud-based applications.The capacity to deal with both str

137、uctured and unstructured data lets us unlock greater value with generative AI engines,which are exceptional at processing such unstructured data.Enabling many apps,and through them many users,to leverage more data in their work.Going from a linear one-way data flow,where most data would“sink”into st

138、orage only to be used for backwards-looking analytics,if at all,to now a circular,two-way data flow that integrates data with our front-line apps and operations.Because were utilizing our data in more ways,we extract more value from it.And now,more and more use cases for that data are delivered with

139、 AI whether embedded in existing apps,new stand-alone tools,custom“software”that we create ourselves,or service-as-a-software solutions from outsourced providers.Of course,managing this broader data ecosystem and all of the apps and ops activities that are interacting with it has become a challenge

140、of its own.Whereas a previous decade wrestled with big data,we believe the mission of this new AI-powered era is to wrangle“big ops”taming the scale and complexity of all the apps,agents,and automations interacting with this data.Big Data vs.Big OpsBig DataBig OpsScale&complexity of data collected,s

141、tored,and analyzed Scale&complexity of apps and automations interacting with dataSource:chiefmartec&33Martech for 2025Yet for all the advances of the modern data stack,its largely been focused on customer data and business operations data.Far less has happened with content“data”.The Current Personal

142、ization StackThe(Underdeveloped)Universal Content LayerData is Real-Time Unlike Mass Customizing ContentContent Layer(simplified)Data Layer(simplified)Largely integrated&automated processLargely fragmented&manual processContent SourcingCL guide,digital assetsContent Qualitytag,translate,localizeCont

143、ent Modelling?analyse,ranking,etc.Content Activation?content?Contenttest&optimizeDataContentdecide&orchestrateDataHyper PersonalizeData SourcingETL,ingestData Qualitystitch,clean&enrichData Analysismodelling,machine learnData Activationreverse ETLSource:MartechTribe&Rasmus Houlind&34Martech for 2025

144、Yes,we have tools such as digital asset management(DAM)to store brand and campaign assets,such as images and videos,or Product Information Management systems(PIM)to store product information.We have content management systems(CMS),whether tightly integrated with a digital experience platform(DXP)or

145、separated as so-called“headless”CMS.And we have master data management(MDM)platforms to own the definitions for a range of entities and messages in our business.But many of these components have not been well-integrated together.Theyre often connected in a fragile point-to-point fashion to serve spe

146、cific use cases.We only connect them to customer data in the late stages of campaigns or customer experiences to deliver limited personalization.And thats not even considering all of the“content”that businesses have that lives outside of those official repositories,scattered across Google Drives,Wor

147、d documents,internal wikis,and a plethora of indie content creation and curation tools in the cloud,sometimes stored by their agencies.In all fairness,its been harder to organize and leverage all that content across the organization than the comparatively well-structured,API-accessible data associat

148、ed with customers and our business operations.We didnt have the technology to harness this content in more advanced ways,so there was little incentive to invest in a“modern content stack”to the same degree as the modern data stack.Until now.Generative AI has the ability to absorb this vast treasure

149、trove of content and synthesize it into a wide range of creative new use cases.We believe the greatest opportunity is rendering more holistically-personalized customer engagement,where personalized content and experiences seamlessly blend both our deep knowledge about the customer(via the modern dat

150、a stack)and the true“personality”of our brand from relevant content(via a new modern content stack).&35Martech for 2025Analogous to the master files of creative assets from which context-specific derivatives are currently produced,we can envision a“master file”LLM or really a multi-modal model tuned

151、 on the content that represents the brands core identity and personality.This genAI“brand master file”or“brand LLM”orchestrates the universal content layer at its foundation.In turn,there may be campaign-specific variations from this brand model that inherit the core identity and personality but the

152、n fine-tune to the characteristics of the campaign.While such a modern content stack to power hyper-personalization like this is at best nascent today,we expect to see significant innovation in such a universal content layer in 2025.The Future Personalization StackScaling Real-Time Content with GenA

153、I Content Master FilesContent Layer-Largely GenAI(simplified)Data Layer-Largely ML(simplified)Largely parallel process(sourcing,quality,analysis&activation)Largely sequential process(sourcing,quality,analysis&activation)Non customer facingBrandLLMCampaignLLMCollateralLLMContent LLMContenttest&optimi

154、zeDataContentdecide&orchestrateDataHyper PersonalizeData SourcingETL,ingestData Qualitystitch,clean&enrichData Analysismodelling,machine learnData Activationreverse ETLCustomer facingSource:MartechTribe&36Martech for 2025Try to fill in the blank here:“are to AI agents as data is to AI models.”The an

155、swer,if you havent already guessed,is APIs.Data gobs and gobs and gobs of data feed the training of ever more powerful LLMs.OpenAIs GPT-4 model was reportedly trained on more than a petabyte of data.(A petabyte,which is 1,000 terabytes,is roughly the equivalent of 223,000 full-length DVDs.Thats abou

156、t half of all the movies that have ever been made,including classics such as The Empire Strikes Back and not-so-classics such as Jay&Silent Bob Strike Back.)Data powers AI models.AI agents leverage those data-powered AI models,but they add a powerful capability on top of them:the ability to use tool

157、s and take actions.While AI models can“talk”the talk,suggesting actions for us humans to go do,AI agents can actually“walk”the walk and take actions directly themselves.This action-orientation is resulting in a new generation of AI models called LAMs,large action models,a label that rhymes with LLMs

158、,large language models.Cobus Greyling produced a useful summary of the differences between them:API Composability as AI Agent Building Blocks&37Martech for 2025FeatureLarge Language Model(LLM)Large Action Model(LAM)Primary FunctionProcesses and generates natural language(text)Executes actions and de

159、cisions in real-world or simulated tasksCore TaskText generation,language understanding,answering queriesPerforming tasks,interacting with tools,and decision-makingExamplesGPT-4,BERT,T5xLAM,AlphaGo,ReActKey ApplicationsChatbots,translation,summarization,content generationAutonomous agents,robotics c

160、ontrol,task completionTraining DataMassive text corpora(e.g.,books,websites,documents)Specialized datasets that include actions,outcomes,or decisionsArchitecture FocusText processing,sequence prediction,context understandingAction-oriented,combining reasoning&decision-making strategiesOutput TypeNat

161、ural language text(e.g.,paragraphs,responses)Actions,decisions,or task completions(e.g.,API calls,moves)Example Use CaseWriting an article,answering questions in a customer support botControlling a robot to stack blocks autonomouslyExample Use CaseSummarizing scientific research papersGenerating the

162、 correct sequence of steps to complete a function-callInteraction with ToolsLimited,requires fine-tuning or integration for specific tasksDesigned to handle tool usage and decision-making in real timePerformance BenchmarksNLP benchmarks like SuperGLUE,SQuADAgent performance benchmarks like Berkeley

163、Function-CallingGeneralizabilityHighly general,handles a wide variety of language-based tasksTask-specific,often fine-tuned for specific domains or toolsetsIllustrationLLM:A chatbot providing movie recommendations based on reviewsLAM:A home assistant autonomously adjusting lights based on user prefe

164、rences and current conditionsLarge Language Models&Large Action ModelsSource:https:/ for 2025How do the LAMs and AI agents actually take actions?Some will simulate human use of a computer mouse movements,button clicks,text input,and“reading”the output.This is what Anthropics latest Claude 3.5 Sonnet

165、 release in October enabled.Its kind of a next-generation approach to robotic process automation(RPA).But simulating humans is messy and computationally expensive compared to letting AI agents work in their native digital tongue by calling APIs.As described earlier,AI agents can generate small softw

166、are programs on-the-fly behind the scenes to take many of their actions.For software programs,the easiest and most direct way to get things done is to call the APIs of other software programs.A recent report by Workato,Behind the Hype:The 2024 State of LLMs in Business Processes,revealed that nearly

167、 half of business leaders(n=1,000 in North America)see APIs as the best way to leverage generative AI in their business processes significantly more than through conversation chat interfaces or RPA-like user interface simulations.APIs have the benefit of being well-defined,versioned,predictable in t

168、he format of their input and output,and able to be accessed and paid for in a more granular fashion.In the future,which method to use generative AI in business processes do you believe will be best for your organization?0%100%21%18%9%29%13%6%4%29%Manual Via Chat23%Other48%Via APICustom code via APII

169、ntegration/automation platform via APIOtherRPASaaS AI featuresSpecialized AI orchestration via APIManual via chatSource:The 2024 State of LLMs in Business Processes Report,WorkatoNote:n1,000&39Martech for 2025This is why APIs are to AI agents what data is to AI models.The good news is that were alre

170、ady on the road to support agents with APIs by leaning into increased“composability”in our tech stacks.Back in May,we published the The State of Martech 2024 report,which did a deep dive into the rise of composability in martech stacks.We know,“composable”has become a buzzword in martech these past

171、few years,with composable CDPs,composable DXPs,and such.But the essence of composability has been around in technology for decades.Its the ability to take different pieces from your tech environment data from multiple sources,APIs from multiple platforms and services and combine them together into n

172、ew workflows or customer experiences.In a very real sense,all software is composed,using programming frameworks and code libraries.But for many years,composing new digital creations required technical skills.Degrees of ComposabilityComposed stackComposed datasetComposed workflowComposed app/websiteC

173、omposed StackComposed capabilitiesComposed experiencesComposable Platforms&ProductsComposed CreationsCustomer experience or employee experienceSource:chiefmartec&40Martech for 2025One of the powerful effects of AI is a shifting of the skills required to use such composability to empower less technic

174、al users.More and more composition of workflows and experiences can be achieved through visual interfaces or natural language requests to AI co-pilots and agents.This in turn has made APIs and open data models more important to martech buyers.In our State of Martech 2024 survey this past spring,83.9

175、%of the participants said APIs were important or very important when evaluating martech products.With the rise of AI agents and automation in the year ahead,this will be more true than ever.However,while most martech products have certainly improved their API coverage over the past few years,only 17

176、.3%of the platforms marketers have at the center of their stack are rated as having great API coverage,enabling marketing ops to do everything they want via APIs.Spectrum of ComposabilityServicesNon-technicalTechnicalDataCRM platformsLow-code app builders(e.g.,Power Apps)Code librariesCloud API serv

177、ices(e.g.,AWS)Composable DXPsNo-code website builders(e.g.,Wix)No-code autom.tools(e.g.,Zapier)Spreadsheets(e.g.,Excel,Google Sheets)Low-code ETL tools(e.g.,Fivetran)CDPsCloud data warehouses(e.g.,Snowflake)BI&analytics(e.g.,Looker)No-code databases(e.g.,Airtable)Enterprise automation tools(e.g.,Wor

178、kato)Shifting with AISource:chiefmartec Note:Products mentioned above are only approximate examples&41Martech for 2025How important are APIs when youre evaluating a new martech product?How complete are the APIs for that platform at the center of your stack?Source:2024 Martech Composability Survey,ch

179、iefmartec&MartechTribeSource:2024 Martech Composability Survey,chiefmartec&MartechTribe51.2%32.7%11.3%4.8%Very important one of the top requirementImportant part of our core requirementsSomewhat important a secondary criteriaNot important83.9%17.3%47.6%28.6%5.3%1.2%Great coverage we can do everythin

180、g we want via APIsGood coverage we can do most of the things with want via APIsOkay coverage we can do some of the things we want via APIsPoor coverage we can do very few of the things we want via APIsNot applicable we dont use APIs&42Martech for 2025Our advice to martech vendors:seize this opportun

181、ity for competitive advantage.Having better APIs for your product is a way to distinguish yourself from your competitors.You can enable your customers to do more automations and soon more AI agent magic by providing great programmatic interfaces to your products functionality.Because increasingly,yo

182、ur users arent just going to be humans.Theyre going to be automations and agents operating on behalf of humans.But just as your closed/lost or churn risk increases if a human user cant work with your product,you can expect the same risk will rise if their“artificial users”cant either.Now,some martec

183、h vendors might be worried that enabling greater API access to their functionality will reduce the usage of their human-oriented UI.Since human attention is extremely limited and highly valuable,there is an incentive to want to catch and hold on to as much of it as possible.It might be tempting to s

184、ay,“No,you can only unlock the power of our product by giving us your full attention as a human in our UI.”For most products,we think that would be a mistake of Innovators Dilemma magnitude.Yes,human attention is valuable.But its not the only source of value for a vendor.As an army of AI agents thun

185、ders across the field,those vendors who are able to empower such artificial users will have greater usage than those who dont.In the Age of AI,usage will be even more important than UI.Increasingly,this will be how most martech products monetize.The Age of AI will also be the Age of APIs,and we expe

186、ct to see that play out in practice in martech stacks over the year ahead.&43Martech for 2025Generative AI is moving faster than any other technology weve seen before:the pace by which its foundational capabilities are evolving,the speed by which those capabilities are being added to new and existin

187、g products across the martech landscape,and the rate by which marketers are adopting or at least experimenting with these new capabilities in their work.Adoption curves used to be measured in years.With generative AI,adoption is growing in months or even weeks.Among the many challenges this presents

188、,its hard to report the real adoption of genAI at any given moment in time.Its like the Heisenberg uncertainty principle in quantum mechanics,a trade-off between measuring the position vs.the velocity of a particle.In our case,the deeper and more detailed one looks at how genAI is being used at any

189、one time(position),the more likely that data will already be out of date by the time the survey is completed(velocity).This has led most surveys of genAI adoption to stay fairly high-level,e.g.,is it being used in any context in marketing and sales vs.IT,legal,finance,HR,etc.,to show that larger tre

190、nd over time.We decided to go in the other direction and dig into the more granular use cases in marketing and martech.In October,we completed one of the most in-depth surveys of how marketers are currently using generative AI across different martech categories and product segments.While this data

191、is more likely to change quickly with high variance,we believe it is directionally relevant for understanding genAI adoption in martech headed into 2025 adoption across all these martech categories is only going to grow.Inclusive of a diverse mix of B2B and B2C industries,we surveyed 283 respondents

192、 globally.3.How Marketers Are Using Generative AI Today&44Martech for 2025Lets start with 50+popular use cases across 49 martech subcategories where generative AI is being adopted:Most Used GenAI Use CasesPopular Martech GenAI Use Cases0%20%40%60%80%100%Content-Copy IdeationContent-Copy ProductionMa

193、nagement-Transcription,Notes,SummariesContent-Content Optimization&TestingManagement-Content IdeationContent-PersonalizationContent-Image/Video IdeationData-Knowledge&DocumentationManagement-Knowledge&DocumentationData-Chat with Data&InsightsData-Competitor ResearchContent-Image/Video ProductionCont

194、ent-Email Analysis&DeliverabilityData-Data Extraction&ConversionContent-Website/Page BuildingData-Coding&DevelopmentData-DashboardsSocial-Content IdeationContent-Social Media ManagementData-Data SourcingSocial-Content ProductionData-Audience BuildingContent-Lead Scoring&DistributionSocial-Data Insig

195、htsAds-Ads IdeationSales-Sales ContentSales-Sales Support&Meeting NotesSales-Personalization&OptimizationAds-(Contextual)AdsSocial-ChatbotData-Data IntegrationSales-Pitch DecksContent-Audio/Podcasts ProductionSocial-Social Media Analysis&Mgt.Ads-Ads ProductionAds-Media/Ads Mgt.Social-Documentation P

196、roductionSales-Playbook/Next Best ActionSales-Lead Sourcing&OutreachManagement-Sales AssistantsSocial-Customer Service&SupportAds-Video Ads CreationAds-Social Media AssistantsSales-Pipeline OptimizationSocial-Community&Review AnalysisManagement-Talent Management/RecruitmentManagement-ComplianceData-

197、Compliance&RiskAds-Brand SafetySocial-Shop/Product Assistants69%62%53%49%46%45%44%43%43%39%38%36%33%33%30%29%29%28%26%26%25%25%24%24%23%22%22%22%21%21%21%20%18%18%18%18%18%16%16%16%16%15%15%13%12%12%11%11%8%7%Source:2024 GenAI Survey,chiefmartec&MartechTribe&45Martech for 2025Not surprisingly,conten

198、t ideation(69%)and content copy production(62%)were the two most popular use cases.Those have been the easiest ways for people to get started with tools such as ChatGPT,Claude,and Gemini,as well as early pre-ChatGPT products such as Jasper and Copy.ai.Many of the other Top 12 use cases are related t

199、o accelerating marketings content development pipeline:content optimization and testing(49%),broader content ideation(46%),image/video content ideation(44%),and image/video content production(36%).Content ideation and production are 7 out of the top 12 use cases in marketingThe#3 use case transcript

200、ion,notes,and summarization of meetings(53%)was also one that predated ChatGPT.Zoom launched meeting transcriptions in 2018.Companies like Gong built entire businesses around the underlying capability to quickly and cheaply transcribe calls.The cost of this capability has plummeted,making it univers

201、ally available.Not just for calls,but summaries of support tickets,deal notes,email threads,etc.As genAI has exploded the amount of content in the world,the parallel ability to reduce that flood of content down to its essence and key 0%20%40%60%80%100%Content-Copy IdeationContent-Copy ProductionMana

202、gement-Transcription,Notes,SummariesContent-Content Optimization&TestingManagement-Content IdeationContent-PersonalizationContent-Image/Video IdeationData-Knowledge&DocumentationManagement-Knowledge&DocumentationData-Chat with Data&InsightsData-Competitor ResearchContent-Image/Video Production69%62%

203、53%49%46%45%44%43%43%39%38%36%&46Martech for 2025The next most popular cluster of use cases which kind of includes such summarization is using generative AI to more easily find and consume data.Better search functionality across databases,documents,and knowledge bases(43%each in management tools and

204、 data tools).The ability to engage with analytics by asking questions in natural language,so-called“chat with your data”functionality(39%)is gaining popularity quickly,accelerating the democratization of at least simple,self-service data analysis.Competitive research(38%)has also gotten a boost from

205、 genAI,to both collect and synthesize large quantities of competitive content.Adoption of other use cases starts to trail off from there.But theres still a tremendous amount of experimentation with genAI in different martech tools,albeit distributed across more categories.Empowering marketers to mor

206、e easily find and consume more data and information are 5 of the top 12 use casespoints for quick human consumption has proven to be extremely useful and appreciated.(This summarization capability also poses a challenge to marketers as their target prospects and customers start using it to collapse

207、marketings perfectly crafted emails and web pages into short bulleted highlights.)0%20%40%60%80%100%Content-Copy IdeationContent-Copy ProductionManagement-Transcription,Notes,SummariesContent-Content Optimization&TestingManagement-Content IdeationContent-PersonalizationContent-Image/Video IdeationDa

208、ta-Knowledge&DocumentationManagement-Knowledge&DocumentationData-Chat with Data&InsightsData-Competitor ResearchContent-Image/Video Production69%62%53%49%46%45%44%43%43%39%38%36%&47Martech for 2025Ad-related use cases of generative AI were 6 out of the bottom 22We found it a little surprising that a

209、dvertising-related use cases showed some of the lowest adoption in this survey 6 out of the bottom 22 categories.We have several hypotheses as to why.We surveyed more marketers than advertising agencies or creative freelancers.Marketers may not have as much visibility into the evolution of ad produc

210、tion that theyve outsourced.Marketers may still be uncomfortable turning over more of the“craft”of advertising to AI.Or,its possible that they are actually using AI without even realizing it(e.g.,Googles Performance Max,Facebooks Advantage+).0%20%40%60%80%100%Ads-(Contextual)AdsSocial-ChatbotData-Da

211、ta IntegrationSales-Pitch DecksContent-Audio/Podcasts ProductionSocial-Social Media Analysis&Mgt.Ads-Ads ProductionAds-Media/Ads Mgt.Social-Documentation ProductionSales-Playbook/Next Best ActionSales-Lead Sourcing&OutreachManagement-Sales AssistantsSocial-Customer Service&SupportAds-Video Ads Creat

212、ionAds-Social Media AssistantsSales-Pipeline OptimizationSocial-Community&Review AnalysisManagement-Talent Management/RecruitmentManagement-ComplianceData-Compliance&RiskAds-Brand SafetySocial-Shop/Product Assistants21%21%21%20%18%18%18%18%18%16%16%16%16%15%15%13%12%12%11%11%8%7%&48Martech for 2025W

213、ith all the innovation happening in the advertising space,however take Adobe GenStudio for one we expect those adoption numbers will grow significantly in 2025.If we aggregate respondents using any use case within the six top-level categories of our MartechMap,this is the distribution of adoption of

214、 genAI use cases reported:Content,data,and management use cases are the dominant buckets at least for the audience of marketers who participated in this survey with us.As mentioned earlier,ad agencies would likely index on Ads use cases more,and we expect sales organizations are leaning into Sales a

215、nd Relationships genAI use cases at a higher rate too.GenAI Adoption Agregated by(AIDARI)Category050100150200250AdsContentRelationshipsSalesDataManagement30%79%33%28%61%57%Number of respondents283Source:2024 GenAI Survey,chiefmartec&MartechTribeNote:816 GenAI use cases were mentioned by the 283 resp

216、ondents of this survey.&49Martech for 2025As we covered earlier in this report,one of the interesting tug-of-wars happening within the context of AI is existing martech vendors rapidly embedding AI features in their products and platforms,while brand-new,AI-native startups are popping up by the hund

217、reds to complement or challenge them.We were curious about the relative adoption of new AI tools vs.embedded AI in existing tools for these different generative AI use cases.Overall,while 23%of the usage by respondents was exclusively with new AI tools,and 17%was exclusively with embedded AI in exis

218、ting tools,the majority 60%were using both.Heres a more detailed breakdown by category,ordered by popularity of use cases adopted with embedded AI in existing martech products:How do you use GenAI?New AI Tools vs.AI Embedded in Current Martech ToolsIncumbent PlatformsChallenger PlatformsIndie ToolsS

219、ubsumingDisruptingM&AUse both:60%Use new AI tools:23%Use embedded AI in existing tools:17%Source:chiefmartec&MartechTribe&50Martech for 2025Adoption of AI Tools(embedded tools and/or new tools)Source:2024 GenAI Survey,chiefmartec&MartechTribe0%20%40%60%80%100%Management-Transcription,Notes,Summaries

220、Content-Content Optimization&TestingContent-PersonalizationContent-Email Analysis&DeliverabilityContent-Copy IdeationManagement-Knowledge&DocumentationData-DashboardsContent-Copy ProductionManagement-Content IdeationData-Chat with Data&InsightsData-Data Extraction&ConversionContent-Lead Scoring&Dist

221、ributionData-Knowledge&DocumentationContent-Image/Video ProductionData-Competitor ResearchData-Audience BuildingData-Coding&DevelopmentContent-Image/Video IdeationContent-Website/Page BuildingAds-(Contextual)AdsSocial-ChatbotSocial-Data InsightsSales-Personalization&OptimizationSocial-Social Media A

222、nalysis&Mgt.Social-Content IdeationData-Data SourcingContent-Social Media ManagementSocial-Customer Service&SupportData-Data IntegrationAds-Media/Ads Mgt.Sales-Sales Support&Meeting NotesSales-Sales ContentAds-Ads IdeationSocial-Content ProductionSales-Pitch DecksSales-Playbook/Next Best ActionManag

223、ement-Sales AssistantsSales-Pipeline OptimizationManagement-ComplianceAds-Ads ProductionAds-Social Media AssistantsSales-Lead Sourcing&OutreachData-Compliance&RiskSocial-Community&Review AnalysisSocial-Documentation ProductionManagement-Talent Management/RecruitmentContent-Audio/Podcasts ProductionS

224、ocial-Shop/Product AssistantsAds-Video Ads CreationAds-Brand Safety27%21%20%18%18%17%16%16%15%14%14%13%13%11%11%11%10%10%10%9%9%9%9%9%9%9%9%8%8%8%8%8%7%7%6%6%6%6%6%5%5%5%5%5%5%5%4%4%4%9%10%13%8%19%10%7%19%12%9%6%6%11%8%10%6%9%9%6%4%3%9%8%5%9%5%7%3%6%4%6%8%6%10%6%5%4%5%3%4%5%6%3%3%6%3%4%3%17%16%12%6%

225、30%16%4%27%19%15%12%5%18%17%16%8%9%23%13%6%7%5%5%5%10%11%10%5%6%5%8%7%7%8%8%5%5%3%8%4%5%4%6%4%10%6%4%Embedded in existing toolBothNew AI tool&51Martech for 2025The distribution is similar to the overall popularity of use cases,but with some interesting twists.That the top one is transcription,notes,

226、and summaries(27%using embedded AI,17%using new tools,and 9%using both)makes sense again,widely popular tools before the ChatGPT revolution,such as Zoom,already provided this capability.Large sales engagement platforms,such as HubSpot Sales Hub,Salesloft,Outreach,etc.,also added this capability.The

227、next top two use cases for embedded AI were content optimization&testing(31%using embedded AI,either with or without new tools)and content personalization(33%,embedded and both).The existing martech tools in these categories include CMS,DXP,and marketing automation products that are used to create c

228、ampaign landing pages.And then the next use case was email analysis&deliverability(26%using embedded or both),functionality that is often part of major email service provider(ESP),marketing automation(MAP),and customer engagement platform(CEP)products.Its logical that as CMSs,DXPs,MAPs,ESPs,and CEPs

229、 crikey,talk about an alphabet soup!enhanced their products over the past year with generative AI features that a significant percentage of marketers would start to take advantage of them.In many cases,these genAI powered enhancements slipped naturally into their current workflow in their existing t

230、oolset.The 5th most common use case,copy ideation(37%embedded or both)obviously had a lot of new AI tool adoption ChatGPT.But the fact that it showed up significantly with embedded usage in existing tools is a testament to how easy it has been for current martech vendors to incorporate things like O

231、penAIs GPT-4 API into their product experiences for this sort of creative text generation.The last two in the Top 10 well call out are#7 dashboards(23%embedded or both)and#10“chat with data”(23%embedded or both).These are also categories where marketers have had existing analytics tools that theyve

232、been using,often deeply integrated into their workflows and operations.As those products added new genAI features the most common being the ability to ask for a report or a specific answer from a dataset with a natural language query they were often presented in the UI in a highly discoverable fashi

233、on,encouraging marketers to try them.&52Martech for 2025If we reorder the list to sort by the popularity of using new AI tools indie tools and challenger platforms we see this reshuffling:Adoption of AI Tools(new tools and/or embedded tools)Source:2024 GenAI Survey,chiefmartec&MartechTribe0%20%40%60

234、%80%100%Content-Copy IdeationContent-Copy ProductionContent-Image/Video IdeationManagement-Content IdeationData-Knowledge&DocumentationManagement-Transcription,Notes,SummariesContent-Image/Video ProductionData-Competitor ResearchContent-Content Optimization&TestingManagement-Knowledge&DocumentationD

235、ata-Chat with Data&InsightsContent-Website/Page BuildingContent-PersonalizationData-Data Extraction&ConversionData-Data SourcingSocial-Content IdeationContent-Audio/Podcasts ProductionContent-Social Media ManagementData-Coding&DevelopmentSales-Sales Support&Meeting NotesData-Audience BuildingAds-Ads

236、 ProductionSocial-Content ProductionSales-Pitch DecksAds-Ads IdeationSales-Sales ContentSocial-ChatbotAds-(Contextual)AdsAds-Video Ads CreationSocial-Documentation ProductionContent-Email Analysis&DeliverabilityData-Data IntegrationSocial-Data InsightsSales-Personalization&OptimizationManagement-Sal

237、es AssistantsAds-Media/Ads Mgt.Sales-Playbook/Next Best ActionSales-Lead Sourcing&OutreachContent-Lead Scoring&DistributionSocial-Social Media Analysis&Mgt.Social-Customer Service&SupportData-DashboardsAds-Brand SafetyAds-Social Media AssistantsSocial-Community&Review AnalysisManagement-Talent Manag

238、ement/RecruitmentManagement-ComplianceSales-Pipeline OptimizationData-Compliance&RiskSocial-Shop/Product Assistants30%27%23%19%18%17%17%16%16%16%15%13%12%12%11%10%10%10%9%8%8%8%8%8%7%7%7%6%6%6%6%6%5%5%5%5%5%5%5%5%5%4%4%4%4%4%3%19%19%9%12%11%9%8%10%10%10%9%6%13%6%5%9%4%7%9%6%6%4%10%6%6%8%3%4%3%6%8%6%

239、9%8%4%4%5%6%6%5%3%7%5%3%3%3%5%3%18%16%10%15%13%27%11%11%21%17%14%10%20%14%9%9%4%9%10%8%11%5%7%6%7%8%9%9%4%5%18%8%9%9%6%8%6%5%13%9%8%16%5%5%5%6%6%5%4%EmbeddedBothA specific,New AI tool&53Martech for 2025The top 10 use cases are again mostly about content(6 out of the top 10 categories)and data,but we

240、ighted more towards the production of content,including image and video content.There has been a wave of cool AI image generation and video indie tools over the past two years,which we see in these results(ideation tools at 2.4%,production tools at 2.2%).Top 10 Use Cases Embedded Tools vs.New Indie/

241、Challenger ToolsSource:2024 GenAI Survey,chiefmartec&MartechTribe0%20%40%60%80%100%Content-Copy IdeationContent-Copy ProductionContent-Image/Video IdeationManagement-Content IdeationData-Knowledge&DocumentationManagement-Transcription,Notes,SummariesContent-Image/Video ProductionData-Competitor Rese

242、archContent-Content Optimization&TestingManagement-Knowledge&Documentation30%27%23%19%18%17%17%16%16%16%19%19%9%12%11%9%8%10%10%10%18%16%10%15%13%27%11%11%21%17%Management-Transcription,Notes,SummariesContent-Content Optimization&TestingContent-PersonalizationContent-Email Analysis&DeliverabilityCon

243、tent-Copy IdeationManagement-Knowledge&DocumentationData-DashboardsContent-Copy ProductionManagement-Content IdeationData-Chat with Data&Insights27%21%20%18%18%17%16%16%15%14%9%10%13%8%19%10%7%19%12%9%17%16%12%6%30%16%27%19%15%4%Top 10 Use Cases of New ToolsTop 10 Use Cases of Embedded ToolsEmbedded

244、BothNew AI toolEmbeddedBoth New tool&54Martech for 2025But,as any martech stack manager knows well,adoption isnt binary.An adopted product can be purchased,but not really used;used,by only occasionally;or used frequently,as an integral part of peoples regular work habits.As part of our survey,we ask

245、ed about the frequency of usage of these different categories of genAI tools daily/weekly,monthly,no longer using,or not tried(yet).Here are the results sorted by those most frequently used daily or weekly:Frequency of GenAI Tool Usage by Use Case and CategorySource:2024 GenAI Survey,chiefmartec&Mar

246、techTribeFrequency of GenAI Tool Usage by Use Case0%20%40%60%80%100%Content-Copy IdeationContent-Copy ProductionManagement-Transcription,Notes,Summ.Management-Content IdeationContent-Content Optimization&TestingContent-PersonalizationManagement-Knowledge&DocumentationData-Chat with Data&InsightsData

247、-Knowledge&DocumentationContent-Image/Video IdeationSales-Sales Support&Meeting NotesSocial-Content IdeationSocial-Content ProductionData-Coding&DevelopmentData-Competitor ResearchContent-Email Analysis&DeliverabilityData-Data Extraction&ConversionContent-Image/Video ProductionData-Data SourcingSale

248、s-Personalization&OptimizationSocial-ChatbotSocial-Data InsightsData-DashboardsSales-Sales ContentContent-Social Media ManagementData-Audience BuildingContent-Lead Scoring&DistributionAds-(Contextual)AdsManagement-Sales AssistantsContent-Website/Page BuildingAds-Ads IdeationData-Data IntegrationAds-

249、Media/Ads Mgt.Social-Customer Service&SupportSales-Pitch DecksSales-Lead Sourcing&OutreachAds-Ads ProductionSocial-Social Media Analysis&Mgt.Sales-Playbook/Next Best ActionAds-Social Media AssistantsContent-Audio/Podcasts ProductionSocial-Documentation ProductionSales-Pipeline OptimizationData-Compl

250、iance&RiskAds-Video Ads CreationManagement-ComplianceManagement-Talent Mgt./RecruitmentSocial-Community&Review AnalysisAds-Brand SafetySocial-Shop/Product Assistants51%44%43%31%29%28%26%23%21%19%19%19%18%17%16%16%16%15%15%15%15%15%15%15%14%14%13%12%11%11%11%11%11%11%10%10%10%9%9%8%7%7%7%6%6%5%5%4%4%

251、3%19%18%10%15%20%17%17%17%23%24%3%9%7%12%21%17%17%21%10%7%5%9%14%8%12%11%12%9%5%19%12%10%7%5%10%6%8%9%7%7%11%10%6%4%9%6%7%8%4%4%2%2%2%6%5%2%3%2%2%2%2%No longer usingMonthlyDaily or Weekly&55Martech for 2025For the most part,the top 10 categories of tools used daily or weekly are consistent with thei

252、r overall popularity.However,the next few ranked categories show several notable differences.While using genAI in social media has relatively low adoption overall,those marketers who are using it for ideation or production in that context use it frequently.The same goes for using genAI for coding&de

253、velopment and data sourcing.This is likely reflective of more specific jobs in marketing:social media managers,app and web developers,and data analysts.Not everyone needs those tools,but those who do,use them regularly.Relatively few categories had significant responses of people“no longer using”too

254、ls that they had previously tried to adopt.This shows 0%20%40%60%80%100%Content-Copy IdeationContent-Copy ProductionManagement-Transcription,Notes,Summ.Management-Content IdeationContent-Content Optimization&TestingContent-PersonalizationManagement-Knowledge&DocumentationData-Chat with Data&Insights

255、Data-Knowledge&DocumentationContent-Image/Video IdeationSales-Sales Support&Meeting NotesSocial-Content IdeationSocial-Content ProductionData-Coding&DevelopmentData-Competitor Research51%44%43%31%29%28%26%23%21%19%19%19%18%17%16%19%18%10%15%20%17%17%17%23%24%9%7%12%21%3%6%2%2%1%1%2%1%1%1%1%1%1%1%No

256、longer usingMonthlyDaily or WeeklyTop 15 GenAI Use Cases Most Often Used(daily or weekly)&56Martech for 2025While early versions of genAI image creation tools certainly delivered an initial“wow,cool!”factor the first release of DALL-E,for example they were hard to control for precise design and ofte

257、n produced weird artifacts,e.g.,people with three arms and 7-fingered hands.This likely accounts for their relatively larger tried-but-stopped-using contingents.However,the state of the art with generative AI image generation has advanced significantly over the past year.Midjourney,Adobe Firefly,Flu

258、x,Recraft,and more are now highly capable and regularly used by artists and designers,even if only to rapidly prototype ideas for clients.If its been a while since youve tried.AI generated video is still in its formative stages,but as models such as OpenAIs Sora become available,we expect these use

259、cases will gain considerable adoption in 2025.Finally,if we sort the list by the use cases that marketers have not tried yet,we get a sense of where new opportunities to apply generative AI exist:how sticky genAI is,even on such an accelerated adoption curve.The only two with even noticeable drop-of

260、f were image/video ideation and creation.Source:2024 GenAI Survey,chiefmartec&MartechTribe0%20%40%60%80%100%Content-Image/Video IdeationContent-Image/Video ProductionContent-Lead Scoring&DistributionContent-Copy IdeationContent-Copy Production6%5%3%2%2%Top 5 GenAI Use Cases Martekers Used but Abando

261、ned&57Martech for 2025Use Cases That Marketers Have Not Tried(Yet)0%20%40%60%80%100%Content-Audio/Podcasts ProductionData-Compliance&RiskContent-Social Media ManagementContent-Lead Scoring&DistributionManagement-Talent Mgt./RecruitmentContent-Email Analysis&DeliverabilityManagement-ComplianceContent

262、-Website/Page BuildingManagement-Sales AssistantsData-Data IntegrationData-Audience BuildingContent-Image/Video ProductionData-Data SourcingContent-PersonalizationData-DashboardsData-Coding&DevelopmentContent-Content Optimization&TestingData-Data Extraction&ConversionContent-Image/Video IdeationSoci

263、al-Shop/Product AssistantsData-Competitor ResearchData-Chat with Data&InsightsSocial-Community&Review AnalysisAds-Brand SafetySocial-Customer Service&SupportData-Knowledge&DocumentationSales-Pipeline OptimizationAds-Video Ads CreationSocial-Documentation ProductionAds-Social Media AssistantsManageme

264、nt-Knowledge&DocumentationSocial-Social Media Analysis&Mgt.Sales-Playbook/Next Best ActionSales-Lead Sourcing&OutreachAds-Ads ProductionContent-Copy ProductionSocial-ChatbotSocial-Data InsightsManagement-Content IdeationAds-Media/Ads Mgt.Sales-Pitch DecksAds-(Contextual)AdsSales-Personalization&Opti

265、mizationAds-Ads IdeationSales-Sales Support&Meeting NotesSocial-Content ProductionSales-Sales ContentContent-Copy IdeationManagement-Transcription,Notes,Summ.Social-Content Ideation51%46%43%43%41%40%40%38%36%34%31%29%29%26%26%25%24%24%23%23%19%19%19%18%15%15%13%13%12%12%11%11%10%9%9%9%9%8%8%7%7%6%5%

266、5%5%4%4%4%3%3%Source:2024 GenAI Survey,chiefmartec&MartechTribe&58Martech for 2025The top yet-to-be-tried use case,audio/podcast production,is somewhat surprising,as there are a number of excellent genAI tools that turn text into high-quality audio:ElevenLabs,Murf,and WellSaid to name a few.This is

267、a terrific way to repurpose or remix existing text content such as blog posts into multi-modal content.Its an easy way to generate voice overlays.And it also works well for producing translated versions of podcasts and audio tracks.If you havent yet tried Googles NotebookLM to turn large amounts of

268、content blog posts,PDFs,YouTube videos,whole websites,Google Docs,etc.into snappy podcasts with multiple synthetic(but realistic sounding!)hosts discussing the key points and perspectives within,you should check it out immediately.Beyond“chat with your data”,this is more like listening to your data

269、chat with itself.Fascinating.&59Martech for 2025We also asked respondents if they had a policy in place for how generative AI could/should be used within their organization.The results were about 50/50 48%had no policy in place,leaving individual teams free to decide acceptable uses.A bare majority,

270、52%,had a genAI policy.Most of those with a policy had it set by a centralized team that spanned departments across the organization.A small number had a policy controlled by a decentralized“AI labs”function that limited AI to a separate team experimenting in a more sandboxed fashion.The presence or

271、 absence of a policy showed material effects on genAI use case adoption.For example,when we asked people why they stopped using generative AI tools,the most common answer other than the maddeningly vague“Other”was that it didnt save time(35%).Compliance reasons for stopping were cited only 12%overal

272、l.Do you have a GenAI policy?Generative AI Policy and Impact on Usage0%20%40%60%80%100%No,Its free for each teamYes,via a central teamYes,(decentralized)Labs.48%44%8%Source:2024 GenAI Survey,chiefmartec&MartechTribe&60Martech for 2025But when we segment the answers to why people would stop using gen

273、erative AI tools based on having or not having a genAI policy,some interesting differences appear.Naturally,stopping for compliance reasons had a substantial delta:8%for those without a policy vs.19%for those with one.More interesting was that those with a genAI policy were more likely to stop using

274、 a tool because it was too expensive(14%vs.3%for those without a policy),suggesting that a governance framework helped raise awareness and/or enforced guidelines around the cost of genAI usage.Why did you stop using GenAI tools?And do you have a GenAI policy?0%20%40%60%80%100%Didnt save timeComplian

275、ce ReasonsToo expensiveOther35%12%6%47%0%20%40%60%80%100%Didnt save timeCompliance reasonsToo expensiveOther33%19%14%35%40%8%3%49%With policyWithout policySource:2024 GenAI Survey,chiefmartec&MartechTribeSource:2024 GenAI Survey,chiefmartec&MartechTribe&61Martech for 2025Those with a policy were les

276、s likely to give the indeterminate“other”reason for stopping usage as well 35%with vs.49%without also signaling a sharper understanding of genAI and when to use it.That explanation of greater AI maturity is also bolstered by the data that those with a policy were less likely to say a use case wasnt

277、saving them time:33%with vs.40%without.One other difference worth noting:those with a policy were more likely to use genAI embedded in existing martech products and platforms(43%with vs.32%without).Inversely,those without a policy were more likely to use stand-alone,specific AI tools(43%without vs.3

278、0%with).It would make sense that those without a formal generative AI policy yet are still experimenting with isolated and/or inexpensive tools theres no shortage of free generative AI tools on the Internet.As a company matures their AI strategy and operations,theyd be more likely to adopt genAI cap

279、abilities in their primary martech products and platforms,with clearer governance.What GenAI type are you using?Do you have a GenAI Policy?0%20%40%60%80%100%Embedded in existing toolsA specific,new AI toolBoth43%30%27%32%43%25%With policyWithout policySource:2024 GenAI Survey,chiefmartec&MartechTrib

280、e&62Martech for 2025Which industry does your company operate in?Demographics of Survey Respondents38%Technology 25%Professional Services10%Telecommunications,Communications,&Media5%Pharmacy,Healthcare,&Life Sciences4%Banking&Financial Services4%Retail3%Public sector,Government,&Education3%Manufactur

281、ing3%Travel&Hospitality2%Building&Construction 2%Consumer Products 1%Automotive 1%Charity 1%Energy&Utilities 0.5%Industrial&Chemical 0.5%Transportation&Logistics100%Source:2024 GenAI Survey,chiefmartec&MartechTribe&63Martech for 2025What is your role?100%17%Chief Executive Officer16%(External)Projec

282、t Manager/Consultant14%Marketing Director10%Marketing Operations Manager,MRM Manager10%VP,SVP or EVP of Marketing7%Marketing Manager5%Sales Manager4%IT Manager4%Marketing Executive/Coordinator8%Chief Marketing Officer4%Program Manager Operational Excellence1%Creative Director 1%Marketing Procurement

283、 ManagerSource:2024 GenAI Survey,chiefmartec&MartechTribe&64Martech for 2025A conversation with Chris ONeill,CEO of GrowthLoop.The following is an edited transcript of our discussion.Chris,were so excited youre joining us.You took over as CEO of GrowthLoop earlier this year.Could you start by tellin

284、g us a bit about GrowthLoop and your background?Youve done fascinating work in this industry from a number of different angles.GrowthLoop is a data and AI platform that helps enterprise teams get more value out of their customer data by launching highly targeted,impactful campaigns directly from the

285、ir data cloud.Our tool helps teams ingest,organize,and activate data across any marketing or sales destination.After re-ingesting campaign results back into the cloud,our AI engine can analyze those results and make recommendations on how to improve future campaigns.We call this The Loop,and it help

286、s marketing teams launch faster,smarter,and more powerful campaigns that ultimately drive real business outcomes.In terms of my background.Ive been in this game for a while.I was really fortunate to be at Google for about 10 years,from 2005 to 2015.4.Five Perspectives on Martech for 2025End-to-End M

287、arketing on Your Data Cloud with AI&65Martech for 2025Ive had a variety of roles since,leading teams at other tech companies including Glean,Xero,and Evernote.I also serve on the board of Gap,given my interest and background in retail.It was a really nice run at Google,and Im happy to be back in the

288、 marketing technology world.I found myself back in this space because of the GrowthLoop team and the promise of what artificial intelligence can unlock for enterprise marketers.During my time at Google,martech was essentially born.I witnessed Google go from kind of an afterthought to being,of course

289、,an important part of every marketers arsenal.It was also the early days of machine learning,and we started to see what it could really do for the industry.I see the same promise now with AI and its incredible potential to transform our field.Similar to that time period when you were leading Google

290、Canada,and martech was taking off,were now in another major wave of innovation and transformation with AI.How similar or different is it this time?And what lessons should we bring with us?That time at Google was really a story of movement in media,from analog to digital.If you think about artificial

291、 intelligence,its not a new concept.Its been around for a long time.In fact,it was coined in the 70s at my alma mater,Dartmouth College.There are three required conditions for any big leap in artificial intelligence.1.The availability of near-or real-time,high-quality data.2.Task-specific algorithms

292、3.Courage.It takes courage from practitioners to say,“Theres a better way to do things,and were going to try it.”So,if you think about Google in the early 2000s and 2010s,it was applying machine learning and artificial intelligence to advertising.Historically,media mix was determined by gut and feel

293、.It was much simpler then,but as channels proliferated and we had actual data to&66Martech for 2025quantify the impact,that all changed.You had the availability of real-time data,both on the demand side and the supply side for advertising.You had task-specific algorithms that would measure if we ser

294、ved ads,and if people responded,clicked through,or took an action downstream.And then you had a lot of courageous people.Those were the folks I enjoyed working with the most:courageous,ambitious people who were bold enough to try something new by leaning into digital instead of newspapers and TV.Tho

295、se conditions have existed,whether youre talking about Moneyball with baseball,AlphaGo,or self-driving cars.It comes down to the availability of data,task-specific algorithms,and courageous practitioners who push the edge of whats possible.So thats the backdrop and similarities of what happened at G

296、oogle back then.So,whats different,whats changed,and why are we seeing so much excitement?Why am I so excited?There are three driving forces:The first goes back to the growth and adoption of the cloud data warehouse.Over the last 10 years,more teams have seen the value in centralizing their data in

297、platforms like Redshift,Snowflake,BigQuery,or Databricks.Thats huge.The second is the rise of composability in martech,or the ability to have tools that serve a specific function and can plug and play with one another.With a composable tech stack,youre not locked into one system and one vendor.Inste

298、ad,you can mix and match and have best-of-breed tools that fit your specific needs.The third,and arguably the biggest,is the rise of agentic AI the ability to have autonomous or semi-autonomous agents do specific tasks like data ingestion or data scoring.These agents can also manage outcome-specific

299、 tasks,such as building audiences or customer journeys.In other words,name an outcome youd like from a campaign,and the AI can generate the audience or journey that will deliver that outcome.&67Martech for 2025Were starting to see some seismic forces,these tidal waves of innovation happening in the

300、form of data clouds,composability,and an agentic AI boom.Speaking of courage,a category of martech thats had an absolutely meteoric rise over these past five to eight years is composable CDPs.But that category is now going through interesting shifts given whats happening with AI.How do you see the c

301、ategory evolving?I think a lot of change has happened,but at the same time,there hasnt been much change at all.What I mean by that is you had the formation of SaaS and stand-alone systems of record.It started with monolithic systems that each sat within a different area of the business:finance had N

302、etSuite,HR had Workday,and marketing had ExactTarget and Salesforce.But now there are more channels and more cross-channel orchestration opportunities to talk to different customer segments in different ways,meeting each customer where they are.Teams had to bring these systems and data together,not

303、just optimize within a particular silo.This drove the need for a new kind of system of record called a Customer Data Platform,or CDP.Traditional CDP development was very important at that time.Now that seems to be fading in favor of composability and the gravity of the data clouds.Thats what is most

304、 exciting to me.Its not just that the data lives there in the cloud.There are obvious cost advantages,security advantages,and agility advantages to keeping data in fewer places.And thats why were seeing the beginning and continued boom with the data cloud.But now we have the ability to use artificia

305、l intelligence at every step of the way.Agents will be really good at analyzing data and thinking of data schema and those sorts of things.Were already experimenting with that(and seeing success)in our product.Next,how can we orchestrate customer journeys intelligently?How can we extract the right i

306、nformation,get suggestions,and act upon them?How can we figure out the right permutations of the best journeys,determined by the outcome that you care most about:&68Martech for 2025preventing churn,acquiring a new segment,navigating a category slowdown.I see a world where theres outcome orientation

307、and AI agents that serve that objective,in partnership with humans who bring their own level of ingenuity and creativity.I believe these things will come together to create audiences,to orchestrate across surfaces,and most importantly understand whats working so it can,in proverbial“Moneyball”terms,

308、do more of what works and do less of what doesnt.And then start it all over again with more suggestions.Its definitely a compelling picture youre painting.Can we get more concrete about how this is working today?Can you explain“The Loop”and how your customers are using it?Whether you use BigQuery,Sn

309、owflake,Redshift,or Databricks,were native to those cloud environments.Artificial intelligence is going to have a multiplying effect on both the amount and quality of campaigns marketing teams will run using the data in the cloud.Very tactically,you can say,Create some interesting campaigns or audie

310、nce ideas for me.Or you might type&69Martech for 2025very specifically,We are lagging in customer consumer electronics in the West.Come up with some interesting audiences that might help me address this top-level business objective.And out pops,in mere seconds,a whole bunch of audiences built from y

311、our own first-party data in the cloud.Then you can say,“Lets act on this,this,and this.Were going to test this through 5%or 1%holdout to understand what the true incrementality is.”You then set the campaign journey up to send an email,serve up personalized in-app messaging,whatever it is.The journey

312、s can be simple or as elaborate as they need to be to achieve the goal.You launch those campaigns across your channels.And then you re-ingest the performance from the various platforms back into the data warehouse.We have a dashboard,which is basically a lightweight business intelligence tool that r

313、eports in business terms,that details what happened across campaigns.Suppose youre trying to change consumer electronics sales in the West and launched a set of campaigns across various channels.You can use this intelligence tool to see whether you actually moved the needle.Should you do more of the

314、 things that work?Should you shift your approach?The audience creation is where we started just getting marketers unrestricted access to their first-party data for campaign creation.Its the beginning of The Loop.But thinking about how you orchestrate and optimize into the different downstream surfac

315、es is something were training AI on right now by using all sorts of different LLMs.Were earlier in that journey,but have some brave customers,some courageous customers,who are with us on that journey and having fun with it.Lets talk a bit more about this hot subject around AI agents and agentic capa

316、bilities.From your perspective,whats real and what is hype?Well,theres definitely no shortage of hype.Lets be clear about that.But there are a lot of real solutions deployed in market now.For example,Glean uses agents for internal knowledge discovery.&70Martech for 2025You can index and understand a

317、ll the information that lives inside your company so you can know what your company knows.I helped launch Glean out of stealth and had so much fun.That team is doing really great and Im so proud of them.Theres another company Im involved with called Ema.They use persona-oriented agents to augment ho

318、rizontal workflows tailored to a specific persona or a specific vertical use case.I think youre going to see a lot more agents serving tasks and outcomes.Well start with tasks because thats where the semi-autonomous or the purely autonomous nature of agents can really have an outsized impact.Theres

319、a lot of grunt work or not-sexy work that needs to happen just to make sure data is in the right place.I also see a role for it governing data processes,making sure that agents are behaving and that your governance models are being enforced.Thats where Im seeing interesting players think through tas

320、k orientation and then outcome orientation.So were creating audiences that way now.And then were working on journeys and decisioning and campaigns.Then the real magic starts when you can optimize across all of those different outcomes with a higher-order business outcome in mind,such as churn or cus

321、tomer lifetime value.As Bezos wrote about Amazon,a companys job is to maximize its intrinsic value,which is really the discounted cash flows.If you want to roll all the way up to enterprise value,thats how I see the full game being played.AI is now added on top of so many other innovations that mark

322、eters already have to deal with.So its almost like a plane that you need to upgrade while flying it.Do you have any tips and tricks,ideas,or observations around this?My advice:it has to start with the data.The expression weve used for a long time that came straight from our CTO,Tameem Iftikhar,and i

323、s starting to take hold:you dont have an AI strategy if you dont have a data strategy.You have to get the&71Martech for 2025data in the right place.You have to have governance.Without that,the rest doesnt really matter.So I think that thats where it starts.Thats my first bit of advice.I also think t

324、here needs to be a shift away from splashy demos to more of a focus on outcomes.You cant test for wow factors.I think thats the phase weve been in with AI you got me with the“wow”.But how do you show me the“how”and the“now”of the impact of AI?Thats the era that were moving into.People need to be cle

325、ar about the outcomes theyre trying to achieve.What is a specific business problem youre seeking to solve and utility youre looking to get from these models or tools?And then really demand to know how the AI tool can scale.Its one thing to do a fun one-off demo.But if it doesnt scale to meet your co

326、mpliance and your data and your governance requirements,then it just doesnt work.Its not worth it.So thats where I think we are right now.Were in that“show me”phase.Earlier you mentioned good data and courage.We love that youre combining technology on the one hand and human factors on the other.What

327、 organizational changes are going to be necessary for businesses to unlock the potential here?Yeah,I think a bunch about this.When we engage in conversations with senior executives,Id say the wiser ones ask this question:How do I bring about change?First,Ill come back to acknowledge that the data as

328、sets themselves are incredibly valuable and you have to invest in them and remove potential silos or barriers that exist.The data has to flow and there has to be clarity of ownership over that.You have to assign clear accountability.This game begins and ends around the data.Then I encourage cross-fu

329、nctional behavior.So if you think of Amazons two-pizza team concept,part of the beauty of that model&72Martech for 2025is that they bring a designer,an analyst,a product person,and an engineer together as a small team to build things quickly in iterative cycles.This concept of agile software develop

330、ment also applies to business.Thereve been times in my career when Ive used agile for everything down to legal queues.Its not perfect,but its surprisingly robust.What Im really getting after and hoping to coach people through is to take an outcome work-back orientation.And then adopt an agile softwa

331、re development mindset to pursue those outcomes,knowing that you may not get it right the first time.Part of what we talk about with The Loop is we want people to get through their first Loop as quickly as possible,even if the learnings or the lift isnt as dramatic as they might hope.But the expecta

332、tion is that by going through and learning,the results will compound.Youre going to do more experiments and youre going to learn faster.Its iterative in nature.And the beauty of these tools is theyre cost effective in that iteration.You go through loops from hypothesis all the way through to outcome

333、 very quickly.The last thing Id say to business leaders is to encourage a culture where it is okay to fail.When youre doing this,the courage part is creating time and space to allow people to experiment so they can find better ways.Often theyre not Eureka things.Theyre compounding little things.Thats the nature of this.So the mindset isnt,“Eureka were going to all of a sudden unearth some silver b