System Selection Isn't About Features Anymore

I recently explored rebuilding the CDP Institute’s venerable RFP Generator with a vibe coding platform. The project was a success – Replit* took about one hour** to build and deploy an attractive, bug-free system that is significantly more capable than the original.***

But looking closely at the RFP Generator – which has been running with virtually no maintenance for six years – also raised the obvious question of how it should be updated. The current focus of the system is features: users specify their target use cases and the system returns a list of the features those cases require; it then suggests vendors based on how many of those features the vendors provide. That made sense when most systems were largely self-contained bundles of capabilities. Features were the main differentiators and collecting information on each system’s features was the main task of a vendor selection project. Even hands-on evaluation tools like pilot projects and bake-offs, which have to some extent replaced or at least supplemented the traditional Request for Proposal, are still largely aimed at understanding system features more clearly.

But systems today are anything but self-contained. They are components of a larger architecture, more-or-less-fluidly exchanging data and participating in workflows that span different systems, departments and even companies. Any new software has to fit into this larger ecosystem.  

This shifts the focus of the purchase process to integration and compatibility. Understanding those is important because different systems are built to play different roles: some are designed to be a company’s primary platform (such as a data warehouse or marketing suite), some are designed to supplement a specific other platform (such as applications on a vendor’s app exchange or an optional module in a platform system), and still others are designed to work with any system that supports the same integration methods.

This means that delivering the right features is no longer enough. Systems must deliver the right features in a way that fits with the company’s larger system architecture. Compatibility with that architecture is the first screen that must be applied to any potential vendor list. In fact, since new tools often make it practical for users to add their own features to a purchased system, finding exactly the right set of features is considerably less important than it used to be.

The CDP Institute’s existing tools already accommodate this change to some extent. We capture whether a particular CDP works with its own database or connects to an external warehouse, which is one of the key architectural differentiators. We also ask whether a system can be installed on-premises, which is an important consideration for some buyers. The vendor selection portion of the new RFP Generator allows users to filter systems on those options. But we could surely do more to understand the different architectural options and the markers that indicate which systems are compatible with which options, and to collect and expose that information.

It's also important to recognize that new architectural options will continue to emerge. The AI-powered transformation of marketing technology is still in that early stage where the industry is trying many alternatives before settling on a standard approach. (Or several standards.) Once the industry coalesces around a few standards, it will be much easier for buyers to identify systems that match their company standard.  But it will be at least several years before the winners are clear.  This means buyers will struggle for some time to figure out which systems are compatible with their company's chosen approach.

Of course, technology isn’t the only factor that matters when selecting a system. Pricing, industry experience, professional services, typical client size, and local presence are all important considerations. The CDP Institute already captures those in its vendor data and the old RFP Generator already asks the user about them. But it would be helpful to find a way to identify some of the less concrete variables, such as cultural fit, ease of use, and skill requirements.

Ideally, a vendor selection tool would also include a user readiness assessment, to help buyers understand what types of systems are best for them, what kinds of problems they are likely to face, and maybe even what they should do avoid those problems. This is important because -- as I’ve pointed out many times in the past -- by far the most common causes of failure with new systems lie within the purchasing organization, not in the systems themselves.****

In short, finding the right system today is less about finding features and more about architectural and organizational fit. The buying process needs to adjust.  Vendor selection tools should – and I’m confident will – adjust along with it.

_______________________________________________________________________ 

* Lovable also worked very well. Botpress, not so much.

** This was after I spent a day and half prepping the data and writing detailed design instructions. It would have taken much longer if I wasn’t working with materials already assembled for the old system. Still, it would have a taken a conventional developer weeks if not months to deliver a system after they were handed those inputs. Replit cost me $20 in processing credits; conventional developer costs would surely have been in the thousands.

*** You’ll have to trust me that the system is ready to go. There’s no point to releasing it until we make additional changes in the data collected. The same goes for our Vendor Comparison report, which uses the same vendor information as the RFP Generator. Replit built an interactive version of that in just a few minutes. Woot!

****The CDP Institute does have an existing Readiness Assessment tool, although it's limited and not connected with the RFP Generator. 

Can AI Break the Hype Cycle?

Generative AI and its agentic offspring may be foundational technologies, but they are still following the normal technology hype cycle: excitement at the potential, followed by news of failed deployments, followed by a more mature understanding of what’s required for success. The third stage generally comes down to a realization that the main obstacles are organizational, not technical: projects fail due to poor internal alignment, change management, reward systems, and so on. Similarly, when it comes to customer data (and all other business data), there’s a dawning recognition that whether the data resides in a CDP, cloud warehouse, or marketing platform, the real challenge is still transforming raw data into usable information through quality management, unification, and mapping. 

We’re just beginning to see the third-stage insights from the industry’s more advanced thinkers. It's true that they’re important: the only way people will be convinced to address these issues is by seeing real data and hearing about actual experience. But they are also utterly predictable. 

What’s less predictable is that AI has the potential to remove those barriers. If AI radically changes how work gets done, then the organizational structures designed for existing business processes (and blocking progress) will change as well. For data, AI could – at least in theory – handle preparation tasks that have always been the main roadblocks to deployment. 

These are not subtle ideas and I don’t consider them particularly brilliant insights. But I also don’t see them being discussed (at least explicitly). Instead, most of the AI conversation is still about using AI to replace individual tasks within existing workflows or about creating AI-based workflows within existing organizational structures. What I’m suggesting is a clearer focus on using AI to remove the traditional roadblocks to technology deployment: in addition to redesigning organizations around AI capabilities (with a particular focus on accommodating future change), this might also mean developing AI coaches to help existing organizations with change management. For data management, it means developing AI tools to automate end-to-end data preparation. 

I will somewhat airily leave the details to people who make their living helping organizations with technology management; they will no doubt have many more and better ideas than I can offer here. My goal is simply to convince them to adopt roadblock removal as a conscious objective in their work. This is what will empower AI to deliver the fundamental changes that we all see as its potential.

Fitting Agents into the Sales and Marketing Mix

Much has been written recently about how marketing and sales processes change when human buyers and sellers are replaced by buyer and seller agents: abbreviated, inevitably, as “A2A” marketing. It’s a fascinating topic but just one model that will coexist in the near future with human (or, more precisely, non-agentic) buyers interacting with agentic sellers, agentic buyers interacting with human sellers, and, lest we forget, humans interacting with humans. Any consultant will immediately recognize that this cries out for a 2x2 matrix, or perhaps a pair of 2x2 matrices if you want to distinguish business marketing from consumer marketing. For the moment, let’s stick with the single matrix model:

It’s worth making these admittedly-obvious distinctions because each situation raises separate issues, which are otherwise easily jumbled into a confusing heap. Let’s look at each situation in turn.

Human to Human (H2H)

Beyond the literal situation of one seller talking to one buyer, I’d argue this also includes humans interacting with traditional broadcast media, web search, and even non-agent websites. The common thread is that the human buyer does most of the work of asking questions and processing answers. The seller is largely reactive, although there are some situations where she makes choices such as selecting a personalized “next best action”, embedding dynamic content in a website, and setting up conventional search engine optimization. Those choices may be informed by predictive models or some other type of AI, but every step in the workflow is ultimately managed by humans, not agents.

I can’t point to specific data but am pretty sure that H2H interactions still account for the vast majority of today’s sales and marketing activity. This means that marketing and sales teams should still give significant amounts of attention to improving them, even though agentic interactions are vastly more fun to think about. If you absolutely must bring AI and agents into the picture, you can use them behind the scenes to speed up workflows, optimize performance, and analyze results.

Agentic Buyers to Human Sellers (A2H)

This is probably the situation that gets the most attention today. It includes true “buyer agents” (controlled directly by buyers) and “buyer-supporting” agents such as AI search engines and browsers. I call these “buyer-supporting” because they’re not controlled by the buyer, but instead by a company like OpenAI or Google which provides them to buyers at little or no cost.

The distinction matters because companies that offer “buyer-supporting” agents have their own agendas, which don’t necessarily align with the interests of actual buyers. In particular, these companies are increasingly interested in monetizing their products by serving ads within AI search and browser results. Some of these ads will be clearly labeled while others may be subtly embedded in the results themselves. These ads are an opportunity for marketers but may be problematic for users, who could be led to question the objectivity of the AI results.

Concern about biased AI search results could in turn lead to significant interest in true “buyer agents” that consumers pay for themselves. History suggests this will be an uphill battle: as we’ve seen with streaming video, large majorities of consumers typically chose free, ad-supported services over paid, ad-free subscriptions. Still, as streaming video has also shown, a significant fraction of consumers will pay for subscriptions in return for a better experience. This could be a large enough market to support a profitable business. Business buyers are even more likely to purchase agent subscriptions, since they don’t pay with their own money and can easily justify the expense based on better quality results. The precedent here is ad-supported versions of office productivity apps, which have never been broadly successful. There’s a chance that agents could be funded by charging advertisers for access to their owners, although such models have also failed in the past.

Advertising aside, most A2H discussions in martech and adtech circles focus on how sellers can adapt their systems to get the best results from buyer-side agents. This often involves advice on optimizing website design to accommodate search and browser agents, so a given brand receives the best possible treatment. Traditional SEO vendors are frantically expanding their products to meet this need and new AEO (AI Engine Optimization) specialists are also appearing. So far, the solutions are pretty basic: systems run sample queries to measure how often a given brand is mentioned in AI search results and vendors offer design tips to expose the kinds of data that AI agents are looking for. The next level is to look beyond measuring and influencing whether the brand is presented, to how it’s presented in terms of positioning and value. We’ll surely see more of that.

The thing to remember about “buyer-supporting” AI search and browser agents is they are generally driven by a big LLM model that draws from the same information for all users. True “buyer agents” would supplement the more-or-less static LLM models with custom research that visits seller websites to find answers to buyers’ specific questions. For example, one buyer might be interested in pricing details while another cares more about product quality. Beyond exposing all possible information, a seller might aim to present its product differently depending on what appear to be the buyer’s priorities. This is largely similar to today’s (non-agentic) website personalization. What’s more intriguing is the possibility that sellers can find a way to identify individual buyers’ agents over time, perhaps by requiring registration in exchange for detailed information. This would let the seller build a buyer profile and tailor responses to this profile. Piercing the buyer agents’ veil of anonymity would be hugely valuable.

There is a third situation: where the “H” in “A2H” is an actual human, not a non-agentic system. One current example is humans responding to agent-generated Requests for Proposals, which will likely be joined by other formats such as email inquiries or even telephone surveys. The growing volume of agent-generated requests is already a nightmare for business sellers faced with the cost of responding to them. The obvious solution is to let seller agents respond to the buyer agents, but it may be a while before most firms can deploy this capability. In the interim, sellers will be increasingly pressed to qualify buyers before deciding how to respond. Insofar as responding to qualification questions requires effort by the buyer, this imposes a cost on the buyer that should help to eliminate frivolous requests. At some point it might make sense for sellers to impose a literal cost – that is, to charge a fee – for responding to agent-generated sales queries. A less obvious concern is that buyers who rely on agent-generated research questions may fail to understand their true needs, removing a substantial portion of the value gained from a good purchasing project.

Human Buyers to Agentic Sellers (H2A)

Traditional websites may use AI-driven personalization but they are still non-agentic systems. In the future, we can expect true agentic interactions to become increasingly common. The best current example would be chat interfaces connected to an agentic back-end, enabling them to engage in true conversations with potential buyers. These have already evolved in some situations to full-scale agentic business development reps (who send those those super-annoying emails complementing your latest blog post and asking for an appointment) and sales reps (engaging in lengthy dialogs).  Agentic customer support reps are even more common and, often, better than humans at many tasks. While the distinction between AI-based and agent-based interactions can be vague, it’s fair to say that agentic interactions will be significantly more responsive to individual situations. This, in turn, makes them more reliant on capturing real-time data, both for customer behaviors and surrounding context.

Letting autonomous agents interact directly with customers raises major concerns about governance, output quality, and risk. These are widely recognized, as are the challenges of integrating agent-based systems with existing infrastructure. That being the case, I won’t rehash them here, apart from noting that they currently present substantial barriers to adoption of H2A models.

Agentic Buyers to Agentic Sellers (A2A)

Agents selling to other agents is the obvious endpoint of agentic adoption. It’s appealing if only for the amusing prospect of agents merrily jabbering with each other without any human involvement. But apart from a few highly structured interactions, such as programmatic advertising, it’s still largely in the future. A2A can’t become more common until the industry first solves the separate challenges of agentic buyers and agentic sellers. It must then overcome the additional challenges of connecting the two. Once the plumbing issues are addressed, there will be another level of adoption as buyers and sellers work to turn the interactions to their advantage. How will price negotiations work when buyers want the lowest price possible and sellers want the highest price? How will sellers discover the actual needs of buyers so they can make the best recommendations – and is what’s best for the seller necessarily what’s best for the buyer? How will seller agents decide which information to offer and which to exclude? How will agents build trust with each other? And how will companies manage the computing costs of agent-to-agent interactions, which could be substantial if the interactions are extensive?

Plenty of smart people are surely working through these issues. We already see some technical foundations being laid in protocols such as MCP and Google’s A2A. But it’s probably too soon for most marketers to put much energy into worrying about A2A deployment. Mastering the intermediate steps of A2H and H2A should come first and will put them in a better position to deal with A2A when the time is right.

Summary

The impact of AI in general, and agentic AI in particular, is overwhelming. While this piece offers some ideas and makes some prediction, my real goal is much simpler: to suggest that distinguishing the different types of human and agent interactions is a way to split the topic into smaller, more tractable pieces. I hope that helps.

State of Martech Report: Customer Data Platforms Are Evolving, Not Dying

Scott Brinker’s State of Martech report has grown from a one-page logo jigsaw to a small industry, co-authored with MartechTribe’s Frans Riemersma and apparently produced by a small army of elves reviewing thousands of products each year. The latest edition, released yesterday, provides the now-expected deep analysis of industry trends covering category growth, the impact(s) of AI, stack architectures, and more. It’s all interesting and too complicated (or maybe complex?) to recap here, so I can only suggest that you read it on your own. 

In addition to data based on the 15,384 vendors now listed (up 9% from 2024), the report analyzes results from a survey of martech and marketing operations leaders. Again, there’s lots of fascinating information, but I was of course drawn to the sections related to Customer Data Platforms.

On its face, the report offers some pretty bad news for the CDP industry: the fraction of companies citing a CDP as the “center” of their martech stack fell from 15.5% in their similar 2024 survey to 12.5% in 2025. However, the 2025 survey is based on just 96 responses, meaning that’s a swing of three answers, so it’s not cause for too much alarm.* Still, it’s an interesting result, and even more intriguing when you separate B2B respondents (52% of the sample) from others (14% B2C and 34% mixed B2B/B2C). CDPs have never been widely adopted in the B2B world, and indeed, their share is unchanged from 2024 (7.9%) to 2025 (8.0%). The flip side to that is in the B2C and mixed group, the shift is larger: from 26.9% in 2024 to 17.4% in 2025. But, again, bear in mind the sample size: that 17.4% represents seven responses. A shift of three answers is well within the range of statistical noise.

Let’s put sampling issues aside and assume there’s some drop-off in reliance on CDP. The question is, what has taken its place?

Did you just answer “cloud data warehouse, of course”? That’s not entirely wrong – the data warehouse share grew from 20.9% to 23.9%. But the big winner was MAP/CEP (marketing automation/customer experience platform), which grew from 19.4% to 26.1%. CRM grew from 17.9% to 19.5%, or nearly as much as data warehouses. Multi-product suites fell from 1.4% to zero, which hints quite strongly that the respondents heavily skewed away from large enterprises.**

If we combine the MAP/CEP, CRM, and DXP or ecommerce categories into “customer-facing systems”, the combined share of that group grew from 43.3% to 52.1%. So if I were to read any trend from this data, it would be that companies are centering their martech stacks on customer-facing systems, not on data warehouses.

This is actually consistent with the trends we’ve seen in the CDP industry itself, where the most recent major acquisitions (ActionIQ by Uniphore , Lytics by ContentStack, mParticle by Rokt) all involved merging a CDP into a customer-facing product, and where customer-facing vendors like MessageGears, Klayvio, Insider, Listrak, and Braze have added CDP (or CDP-ish) capabilities***. The CDP Institute classifies all of these systems as CDPs, in addition to whatever else they do. So, the way I see it, companies that list those products as the “center” of their martech stack are still building their stack on a CDP, even if they don’t call it one. That’s good news for the industry, not bad.

The survey also takes another look at the role of data warehouses in the martech stack, asking “Do you have a customer data warehouse/lakehouse integrated with your martech stack?” More than half the respondents (56.2%) said they did, a number that climbs to 92% for B2C vendors and 80% for enterprise companies. What’s interesting here is the relation of those answers to the previous question: many more firms have integrated a warehouse than are using their warehouse as the center of their martech stack. This is far from shocking but, again, suggests that the mere presence of a warehouse doesn’t mean that warehouse is the primary customer data store.

If you’re a “warehouse-centric” composable CDP vendor, you could read the relatively small share of warehouse-as-central system either as cause for alarm – the market isn’t as big as you thought – or reason to rejoice – the potential for growth is huge. Both could be true at the same time. But if the primary industry trend is for CDP functions to migrate to customer-facing systems (run by marketing or other business units), then a shift of CDP functions toward the (IT-controlled) data warehouse seems to be the wrong way to bet.  (In this context, the recent sale of leading composable CDP Census to data movement platform Fivetran may signal incipient consolidation in the young-but-already-overcrowded composable CDP sector. That Census was bought by a data movement tool rather than a customer-facing system reinforces the notion that composable CDP products serve IT teams while customer-facing CDPs serve marketers and other end-users.)

Presumably all those non-central warehouses are acting as data sources to CDPs or customer-facing systems with a CDP inside.  Indeed, the Brinker/Riemersma report positions CDPs (stand-alone or embedded) as intermediaries between data systems and activation systems (which they call "systems of knowledge" and "systems of context," respectively), responsible for "organizing or framing the data in a way that best serves more situational needs."  I quite agree, and couldn't have said it better.  I would expand on this by noting that CDPs within customer-facing systems will have direct access to the data those systems generate, so the role of the data warehouse is limited to supplying data that the warehouse collects elsewhere. A CDP with direct access to customer-facing systems overcomes one of the main drawbacks of the warehouse-centric approach, which is that warehouses often can't provide the real-time access to behavior data that's needed for many customer-facing applications of CDP data.

I should stress that CDPs are a bit player in the Brinker/ Riemersma epic. Definitely download the report (it’s free and ungated) and focus on whichever portions you find most relevant. It’s all good.

------------------------------------------------------------------------------------------------------------------------------------

*The 2024 survey had 168 responses, of which 60% were B2B, 9% were B2C and 31% were mixed. Doing that math, that’s 67 responses in the B2C and mixed group, of which 18 cited CDP as the center of their stack.

**The 2025 report says 30 of the 96 responded were from ‘enterprise’ organizations, but doesn’t indicate how these split between B2B, B2C and mixed.

***In addition, at least one unacquired CDP (BlueConic) has repositioned as a customer-facing system.

CDP Round Table: Forget Composable, Are Cloud Databases a Threat to CDPs?

On July 16 and July 18 2024, the CDP Institute hosted a pair of online roundtable discussions for industry vendors. Here are summaries of the conversations.

 

July 16 Roundtable - U.S. and Europe participants.

We started with a quick list of trends that the Institute is watching. These include Composable CDP, CDP integration with advertising channels, third-party cookie deprecation, vertical (industry-specific) CDPs, and AI applications in CDP systems.

The discussion quickly focused on Composable CDP. Key points included:

•  Replacing the term “composable” with “warehouse native”. This is more accurate and resonates in particular with data and IT teams, who want to get the most use from their warehouse investments. “Composable” is still more commonly understood among marketing users.
• Composable CDP comes up most often at large enterprises, which have the most mature IT and data resources.
•  Composable also has appeal in healthcare and financial services, where regulatory concerns make companies reluctant to copy data into a separate CDP database. (Note: I have also heard the opposite, because IT and data teams don’t want to give business users direct access to their entire warehouse and would rather provide them with a limited extract.)
• Digital native companies are also good prospects for Composable because they tend to have well managed data already in place.
•  Composable promises a faster start than building a separate CDP, which is appealing to all potential users. Whether this really happens depends on the state of existing data sources and warehouses.
• Composable CPDs have an advantage because the company’s existing warehouse will include both customer and non-customer data elements tailored to company and its industry. Standard CDPs often must build a custom data model for each new industry, and may struggle to include non-customer information such as product data. Vertical CDPs are appealing because they have industry-specific data models already in place.

Other observations:

• The fast growth of cloud databases, and Snowflake in particular, is a threat to packaged CDPs because they make it easier for IT to build their own products. So far, the cloud databases haven’t built in key tools such as data quality and ID resolution, but there are indications that will change. Such tools are already available as pre-integrated applications in the cloud database vendors’ marketplaces. This is worth calling out as a separate trend from Composable CDP.
• The data engineers who buy cloud databases are not aware that CDP systems exist. They see the value when it’s explained to them, but are still likely to want to expand use of their warehouse systems to justify the investment. This is true even if the warehouse doesn’t already have rich customer data features and customer data has been a low priority. “Data-in-place” and “zero-copy” are strong selling points against putting data into a separate CDP.
•  Regulatory changes, such as anti-redlining rules for banks and Sunshine laws for healthcare marketing, have driven investments in customer data in the past. Privacy regulations may play a similar role in the nature future. This is also worth calling out as a trend.
• There is limited convergence between CDPs and privacy systems. Few CDP vendors have invested in privacy features beyond ingesting consent data, possibly because privacy is complicated so it would take a major investment. In addition, privacy systems have different buyers from CDPs, and privacy managers feel it’s safer to buy from OneTrust than lesser known vendors. But basic privacy and security always requirements in CDP RFPs.
• Journey orchestration is available in many CDPs, which means they overlap with existing journey orchestration systems. But very few clients will replace a mature journey orchestration system with a CDP, due to the effort involved in training staff and migrating programs. Most CDPs with extensive journey orchestration and messaging capabilities are vendors that started as journey orchestration and messaging products and added a CDP capability. Journey orchestration and messaging systems that connect directly with a warehouse may pose another threat to CDPs.
• Advertising integration is often the first CDP application in Italy and elsewhere in Europe.

July 18 Roundtable - APAC participants

Industry trends

• Overview: composable, cloud database, advertising integration, cookie deprecation, privacy & compliance, vertical industry CDPs, AI
• Seeing lots of requests for clean room and consent management, and how CDPs can merge those to streamline work for customers. Some parts of consent can be managed in CDP, others should be in separate platform. Privacy is often first use case.
• Among companies with CDP already deployed, often see extension beyond marketing to other departments. CDP often leads to teams within marketing working together that before kept separate.
• When a new CDP is deployed, people become aware of it organically and through analytic reporting that uses CDP data.
• Cloud databases and composable CDPs introduce new buyer personas from IT and CTOs, who have different use cases and technical concerns from marketers. 

Other items

• Understanding of CDP is relatively low in SouthEast Asia (SEA), compared with Australia, India and Japan.
• Composable is mostly raised by IT teams, who are looking for easiest path to moving data from one platform to another, without understanding the strategy behind the CDP project. Applies to both small and large organizations. Companies like Salesforce with a lot of installed products can co-exist with composable.
• Companies can easily identify many use cases for a CDP. Vendors spend time helping them to prioritize. The sales process is sometimes slowed because buying teams are overwhelmed by the number of use cases.
• Greatest interest among marketers is performance marketing use cases, where results can be directly measured, such as data activation, segmentation, and pipeline-to-paid. This is limiting because there are so many other use cases that don’t have immediate results, such as brand building.
• Advertising is often an early use case, since simple things like suppression and lookalike audiences from CDP can generate immediate, measurable result.
• It’s usually easy to build a use case for CDP looking at almost any part of the lifecycle, from acquisition through churn reduction.
• Confusion about CDP definition can slow down sales, because there are so many different vendors that are hard for buyers to differentiate. The products have changed over time, starting from data capture and now moving to integration with cloud data warehouses and activation in multiple engagement channels, including those outside of marketing.
• It’s not clear whether growth in composable and cloud databases has caused a drop-off in selling stand-alone CDPs. CDP Institute industry update report shows that growth has definitely slowed in the past 18 months, but we don’t know the reason. The slowdown might just mirror the general tech employment slowdown post-covid, and it’s possible that the companies buying composable and cloud databases would have built their own solution anyway, so they were never the companies fueling growth of stand-alone CDP.
• Composable in APAC is having more of an impact in the small to medium sector than with big enterprises.
• AI requires higher data quality.
• Companies feel competitive pressure to invest in AI but are moving slowly, in part due to privacy and security concerns. CDP vendors have added both predictive and generative AI to their systems. Predictive AI deployments are much further along and we’re now seeing initial deployments of generative AI.
• Generative AI is being used to automate existing tasks but not yet to transform tasks. Expect to see more impact next year in terms of benefiting system users, developers, and ultimately at touchpoints where it will change the customer experience itself.
• Generative AI will eliminate some jobs and create others, probably for a net positive impact. People can view AI as a centaur that helps them do their work or a cyborg that takes over their job completely. There are many issues to work out from an enterprise perspective before generative AI becomes widely adopted.