1. What do you see as the defining capabilities of next-generation Customer Data Platforms, and how will they differ from today’s CDPs?

The traditional CDP was designed for a world where marketers needed a system of their own that could collect data, build profiles, and share the profiles. Today, data is largely collected in data warehouses, so the role of the CDP has evolved to focus more on advanced functions for profile assembly – in particular, identity management – and specialized functions for data sharing, such as real-time and reformatting for specific applications.

We also see the scope expanding beyond customer data to include other types, such as product information, and to read data directly from very large and/or fast-changing source systems, such as inventory levels. At the other end of the process, there’s increasing interest in moving beyond data access to having the CDP orchestrate customer journeys through centralized message selection and personalization.

Whether these capabilities are best provided by a single system or by separate components is unclear and probably varies depending on the situation at any particular company. Even when it is a single system, it’s not clear we’ll want to call it a CDP. One thing that hasn’t changed from the original vision is that marketers need as much independence as possible from IT and data teams. No matter how competent they are or how earnestly those teams try, they’ll never move as quickly as marketers can move for themselves.

2. AI is playing an increasingly central role in marketing, with autonomous or agentic systems now emerging. How far are we from seeing AI-driven marketing agents manage campaigns end-to-end, and what challenges need to be overcome for such autonomous marketing to become mainstream?

We are beginning to see end-to-end campaign systems today, such as WPP Media’s Open Intelligence, Google Ads agentic experts, and Meta’s promised ad system. You’ll note that those are all for a single medium—advertising—and they’re run by agencies or publishers rather than advertisers. I expect it will be no more than one or two years before we see systems run by marketers that deliver fully automated campaigns spanning all channels.

Technical challenges include assembling suitable data, giving agents real-time access to that data, getting separate agents to work with each other (addressed by Model Context Protocol), collecting feedback to optimize model outputs, and monitoring to ensure outputs are suitable. Organizational challenges include training staff to manage the systems, cooperation across departments, agreeing on ROI measures, and building enough trust to let the systems run autonomously.

One important challenge I haven’t seen discussed much is the need to predict long-term value rather than immediate response; without this, the agents will optimize on things like clicks.

3. As marketing stacks grow more complex, there is a growing debate about “composable” (warehouse-centric) CDPs versus packaged all-in-one solutions. What is your perspective on this debate, and how should organizations decide which approach is best for their marketing data infrastructure?

The debate around composable systems has died down a bit in the past year or so. I think this is because the industry has recognized that the only valid answer is “it depends”: on the state of a company’s current systems, on their resources to build or integrate components, and on how well available packaged solutions meet their needs. Every company needs to look at those factors and assess for itself which approach will fulfill their current and future requirements.

I’d say the current debate is more between warehouse-centric solutions and solutions embedded in activation systems such as CRM, email delivery, DXP, and ecommerce. So many of those systems now provide significant CDP capabilities that it’s increasingly difficult to justify a separate, stand-alone CDP. The exception is companies with unusually demanding requirements, who are more likely to want a warehouse-based solution that lets them pick or build best-of-breed components.

The traditional stand-alone CDP is increasingly squeezed between those other two options. Again, I’ll note that the warehouse-based solutions tend to be run by IT or data teams, so marketers will still need their own tools to close the agility gaps between what the IT and data teams can deliver and what the marketers really need.

4. How can marketers balance delivering personalized, data-driven experiences with strict data privacy requirements, and what role do CDPs play in ensuring compliance without sacrificing marketing effectiveness?

The answer is one word: carefully. Legal requirements are the simple challenge, since they only define what data is available for which purposes. The greater challenge is deciding when a particular experience will be viewed by the customer as valuable or invasive.

Some general rules of thumb are: don’t use data the customer didn’t knowingly give to you; don’t use data for purposes the customer didn’t intend; and be sure to deliver value to the customer through savings, convenience, recognition, or something else.

You also need trust-building measures as part of your data collection routines, such as explaining how data will and won’t be used; describing your security processes (even if these are no greater than the industry standard); letting customers examine the data you have about them; and offering easily accessible, granular opt-out interfaces.

The role of the CDP is to ensure that only compliant data is made available for marketing programs. This removes the need for marketers to judge what’s permitted on a case-by-case basis. That sounds simple, but in practice often requires elaborate rules taking into account the nature of a data element, how it was collected, the consents granted by each data owner, and regulations which differ by location.

5. What major trends do you see shaping the global MarTech landscape over the next few years, and is there any emerging technology or shift you think marketers might be underestimating or not prepared for?

Obviously the overarching trend is AI, which is already changing both how companies run their marketing and what kinds of marketing are effective.

Other trends not directly related to AI include the growth of interactive touchpoints (in particular, direct purchases through display ads, social media, podcasts, CTV, and other channels), growing privacy constraints, and customer demand for personally tailored treatments.

It’s likely that these trends will lead to MarTech centralization and, perhaps, to simplification as the need to coordinate customer interactions across all touchpoints results in companies settling on a few systems to handle the required functions across all channels, rather than separate systems with overlapping functions for each channel. Marketers may also replace some specialized packaged systems with self-generated AI systems, although I’m still more skeptical that they’ll really be interested or able to do that.

The one change with profound implications that I think marketers are just beginning to recognize is the emergence of AI agents that work on the buyer’s behalf. The initial example is GenAI search, which changes the goal of search marketing from getting a high listing in the search results to getting your product mentioned in the search response. This will require changing how information is presented on the internet, since the GenAI engines look for different kinds of information and formats than traditional search ranking algorithms.

There will be other areas where “buyer bots” will work as agents for customers, and similarly will be looking for different types of information than traditional sales presentations. It’s too soon to know how broadly these sorts of agents will be adopted, but they do have the potential to entirely change what technology marketers need.

6. How do you see marketing technology evolving in emerging markets like India, and what unique challenges or opportunities do marketers in these regions face compared to more mature markets?

Markets like India have an interesting structure: there’s a middle class that is similar to the middle class in developed markets in its attitudes, relative income, and technology level, and there’s a large low-income class that has substantially different needs and resources. The middle class may be a relatively small portion of the population but is still large in absolute numbers in a big country like India, China, or Indonesia.

The technology for marketing to the middle class is similar everywhere, but the technology to market to the low-income segment must meet different needs, regarding their devices, the media they consume, the need for very simple interfaces, the services and products they can afford, etc. This means there is an opportunity to develop MarTech systems optimized for those requirements. There’s a good chance that at least some of these innovations will turn out to be useful in the rest of MarTech since they are likely to be simpler and cheaper than standard solutions.

7. Do you believe that some of today’s biggest AI-powered marketing tools are actually creating more customer distrust than value—because of poor data governance or deceptive personalization?

So far, I don’t think that’s the case. Most AI tools are being used for ideation or analytics, or at most to write copy that is then refined by humans. So poor governance or deceptive personalization are filtered out before they reach the actual consumer.

In general, research shows that consumers are fairly trusting of AI, at least in comparison with non-AI alternatives such as conventional marketing, search engines, or company websites.

8. What advice would you give to marketing leaders trying to navigate the rapidly evolving MarTech landscape? How should they approach decisions around investing in technologies like CDPs and AI to stay ahead of the curve without falling for the hype?

The conventional wisdom for technology management is to adopt a portfolio approach, with most investment in using established technologies to meet current needs, and a smaller fraction set aside for experimenting with cutting-edge systems.

I’d consider CDP to be an established technology, so you should invest when there is a clear understanding of what you’ll use it for and how that will generate value. AI is a newer and more foundational technology. Companies can experiment with small, disconnected AI applications, many of which will save enough labor or improve results enough to justify their cost. These will build some experience with AI but not really prepare you for more profound transformations.

A more important step is department-level implementations that support an entire workflow such as campaign deployment or content generation. This will build more relevant experience without fundamentally changing how your company operates.

The final transformation is end-to-end automation of an entire process such as email marketing or ad campaigns. I’d be careful about that because the technology is advancing so rapidly that any investment is likely to become obsolete in years if not months. Committing prematurely could result in a high-cost, low-performance system that puts you at a disadvantage vs. competitors. Pilot and proof-of-concept projects can be a helpful intermediate step that resolve some uncertainties and expose weaknesses such as inadequate data. Just bear in mind that pilot projects often fail to scale well, so treat these more as experiments and training exercises than as preliminaries to a full deployment.

9. A lot of vendors slap the 'AI' label on basic automation—do you think the MarTech space is becoming another “AI-washing” industry?

That’s more or less inevitable, isn’t it? Marketers need to look beyond the labels to understand what each product can and can’t do, and make sure it meets actual needs before making an investment. Whether it’s “true” AI isn’t important in itself.