The End of the Martech Stack.

There has been a lot of writing about the new martech stack. Scott Brinker, who I have admired for years, has done more than anyone to map and explain the explosion of marketing technology. I had the pleasure of being interviewed by Scott early in his ChiefMartec journey, and I have always respected the clarity he brought to a very messy market. His ideas on the new stack for the AI age are a must read.

Yesterday two more heavyweights Martin Kihn of Salesforce and David Chan of Deloitte chimed in on this idea trying to reckon with what comes next. What is the new stack?

In Silicon Valley there’s been a separate thread about the context graph necessary for AI to thrive. The data. A system of record. Still architectural but increasingly focused on what is required to make AI useful. This is a good starting point for Martech as well.

AI requires its own separate stack that has nothing to do with marketing. This is part of the issue Martech thought leaders are grappling with (and Brinker lays out in his ideas of systems of records and context).

Being in the trenches with enterprise brands, data and AI each day I have my own POV. I think the conversation about the marketing stack is now running into a deeper shift around building AI and as noted above I think it’s why many of the smartest Martech minds are grappling with it. 

Let’s start with the fact that the “marketing stack” is a SaaS construct. It belongs to an older way of thinking. 

The martech stack came from a period when marketers had lots of solutions for lots of problems. Point solutions. The SaaS unbundling. One tool for loyalty. One tool for acquisition. One tool for email. One tool for testing. One tool for personalization. One tool for reporting. 

And these tools could not share data.

That was the world the martech stack was built to describe. It was not built for AI. It was built for software procurement in an era of fragmented applications.

The biggest part of the “Martech Stack” was of course the CDP. Customer Data Platforms. They emerged as a “must-have” because the stack had a data problem. 

Customer data was spread across too many tools. Loyalty data lived in one place. Acquisition data lived somewhere else. Behavioral data sat in another product. Transactional data was somewhere else again. The promise of the CDP was to get that data together and unify it around an entity, namely the customer identity.

In many ways, the CDP was the precursor to this next era. It was an early recognition that data motions and identity were not going to stay stacked inside separate applications. It was going to be aggregated, unified, and organized around the customer. The Customer 360.

And of course there is huge power in that. Even a Customer 180 is light years ahead in terms of what customer insights can be derived.

The more customer data you have, the more intelligence you can create. Why would I want my loyalty data separate from my acquisition data? If those systems are separate, I cannot understand whether the customers I am acquiring are actually becoming loyal customers. I cannot understand how acquisition quality changes by segment. I cannot understand whether customers who come in through one channel have different retention patterns, margin profiles, or lifetime value than customers who come in through another.

***

If the CDP was the precursor to data unification, the data warehouse is the end of that era. The warehouse has become the gravity well / System of Record for customer data: transactional data, behavioral data, enrichments, identity-level data, product data, marketing data, and customer journey data. It’s all there or will be landing there soon.

Once you understand that, you can also see the data warehouse or more likely their lakehouses are exactly where AI can find the data it wants and needs. 

AI does not want a narrow slice of the customer. AI wants email behavior, loyalty activity, media exposure, purchase history. AI wants as much context as possible. It wants horizontal breadth. Tables 1600 features wide. It wants behavioral depth. It wants history. In this data are patterns and relationships between things that could never be visible when the data was trapped in separate SaaS tools.

The warehouse is where customer context is forming from this aggregation and AI for data preparation that exists already. The next generation of marketing AI will be built from that.

To put this marketing centric language, you can’t deliver relevance if you don’t understand context.

So instead of thinking about the AI future in terms of marketing software, let’s think about it from the starting point of the data. 

We have mature and growing systems being built to organize it, govern it, share it, and make it usable. We are seeing more and more multi-cloud architectures. More ability to move data from old systems into new ones. Data motion and data movement across clouds are becoming important precursors to making context building extensible and governed. Any AI system that is going to understand customers, generate strategy, optimize tactics, or improve outcomes has to start from the data. And as much data as possible.

The more horizontal the data view of the customer, the more valuable and actionable the intelligence becomes.

This is where the martech stack starts to break down. The stack was organized around applications with its own data. AI is organized around data and context sharing.

So if the stack is being replaced by intelligence then the question is what kind of intelligence do marketers actually need?

I think it comes down to three things:

Strategy
Tactics
Optimizations

That is what marketers need. This is intelligence. This can all be derived from data. Every part of this relies on context that compounds the closer you get from the data to the customer.

I propose we stop talking about protocols and platforms and start talking in language marketers can understand. CMOs need to know what opportunities exist. They need to know what to do about them. And they need to optimize based on what worked and what didn’t.

That is the job. That is how you grow a business. This is margin creation. This is where AI becomes so important.

AI is going to shorten the learning curve across strategy, tactics and optimizations. With so much more intelligence available, marketers will be able to understand customer needs, motivations, patterns, timing, value, and behavior with much greater efficiency and accuracy.

Better relevance = better ROI. And not just ROI in the media sense. ROI in every sense of the operational word: return on spend, return on time, return on analysis, return on reporting, and return on strategy.

***

Marketing has spent decades trying to infer customer behavior from dashboards, reports, campaigns, and channel metrics. AI changes the center of gravity. It does not just help marketers report on what happened. It helps them understand what is there they haven’t seen. The known knowns and the known unknowns. 

We already have examples of incredible marketing AI technology systems that work this way. PMax from Google and Advantage+ from Meta. In fact, Meta Advantage+ has become a primary driver of Meta’s financial growth, directly contributing to record-breaking revenue in 2026. Somehow it seems to elude people that the new AI “stack” for Marketing has already been built and battle tested at scale no brand alone will ever need.

If we look at how those systems are constructed and use them as a framework for what’s next (the same way so much of the technology these companies developed and open sourced in the past has woven its way into martech, it does not look like traditional martech stacks. 

These systems begin with deep learning. They take massive amounts of data and make sense of it. Using neural networks they find patterns. They organize behavior. They create intelligence. They understand similarity, distance, correlation, affinity, timing, and intent at depths humans cannot get to on their own with a scale of data impossible for humans to understand. 

That is what deep learning neural networks were built to do. And it works!

This is the strategic step. It is about understanding the matching problem. How do people group together? What makes different customers unique? What do they respond to? When do they respond? What do they like? What do they not like? What do they value? What are they likely to do next?

AI allows us to move beyond explicit signals. It allows us to understand latent signals. Not just what someone typed, but what their behavior suggests. What their history implies. What their similarity to other customers reveals. What their relationship to a segment, product, offer, channel, or message might tell us.

That is the strategic layer. 

Downstream at the tactical and optimization level we have reinforcement learning. These RL systems decide what to show people, what offers to present, when to present them, and where to activate them. Do you see a blue sweater or a pink one? Explore vs exploit. These algorithms work great. RL systems can also orchestrate delivery of content with amazing effect. TikTok has proven that. 

That is the tactical and optimization layer.

Still, optimization systems are only as good as the customer context they receive. If the system does not understand the customer well, it can still optimize, but it is optimizing inside a smaller and often shallower view of the world. It is less efficient.

So the real unlock is using these systems together. Replacing the stack with deep learning and reinforcement learning creates a closed-loop system that gets smarter with more data and can be goaled with the objective function of lifting any KPI.

As Sun Tzu said “Strategy without tactics is the slowest route to victory. Tactics without strategy is the noise before defeat.”

In practice the strategic layer is where AI discovers the customer structure, the latent affinities, the patterns, the opportunity pockets, and the context that tactical and optimization AI can use downstream to apply offers, messages, creative, timing, channel decisions, and orchestration and evaluate what worked. Then those results are fed back into the system of record and new strategies and opportunities are discovered again.

This is a system of intelligence. It’s the backbone for the marketing AI that is working now. It will become the backbone of all Marketing AI. But the important shift is that the organizing idea is no longer the stack. The organizing idea is intelligence. Learning. 

And if the old martech stack was arranged around tools and channels, the AI-native marketing system will be arranged around the customer. It will start with data, build context, discover strategy, activate tactics, measure outcomes, learn, and then do it again.

That is the loop.

One last very important point.

That loop is much closer to how marketers actually think. Marketers do not wake up wanting another layer in a stack. They want to know where the growth is. They want to know which customers matter. They want to know what those customers need. They want to know what strategy will move the business. They want to know what actions to take. And they want to know whether those actions worked.

The stack was a useful map for the SaaS era. But AI is not a SaaS construct. AI is a data and learning construct. The companies that understand this will stop asking how to organize more tools. They will start asking how to build systems that understand customers, discover opportunities, apply tactics, and move KPIs.

The future of marketing will not be defined by another stack. It will be defined by systems of intelligence that are always learning.