“Context” is a hot word in martech, but the work required is underestimated

The Princess Bride is one of the best movies ever made. That doesn’t really have anything to do with martech, except to note that our industry has a remarkable penchant for taking a useful word, repeating it incessantly, and stretching its meaning until Inigo Montoya feels obligated to object.

Inconceivable, I know.

Which brings us to martech’s word of the year: context.

Everyone — by which I mostly mean martech vendors and AI influencers on LinkedIn — has been using that word to describe the “special sauce” that makes their agents and agentic machinery work. But what exactly do they mean? That’s where things get a little hand-wavy.

A narrow definition of context would be the instructions, curated data, and permissioned tools that we feed to an AI agent or workflow to perform a task we want done. Hopefully with accuracy and efficiency — and without going rogue and illegally hacking into other people’s systems. (Too soon, AI labs?)

But while that’s functionally correct, it doesn’t explain what exactly goes into that curated context bundle or how to assemble it.

In our State of Martech 2026 report, we stepped back to look at the big picture of what context means strategically:

There’s customer context: the customer’s situation, goals, intent, preferences, history, jobs-to-be-done, and moments-that-matter. And there’s company context: the company’s goals, strategies, knowledge, processes, capabilities, governance, and priorities.

It’s important to acknowledge that these two contexts exist in the real world, regardless of whether we’ve instrumented them — or are even fully aware of them.

That instrumentation is the hard challenge of systems context: how much of customer context and company context is actually represented in our systems in a form that can be acted on. Spoiler alert: typically a relatively small slice. (Note that capturing and storing context data doesn’t necessarily make it actionable.)

The goal of context engineering, writ large, is to expand the coverage of systems context. Meanwhile, the goal of value engineering is to better match the right company context to the right customer context.

The intersection of these three lenses is what we call the Golden Context. Yes, it’s aspirational — much in the same way that the golden record of perfect customer-360 data has been aspirational for, oh, about two decades. But it’s what we should aspire to.

Okay, that’s the 50,000-foot view of context. But what does it mean closer to the ground? Frans Riemersma and I are going to have a ton of research and a practical architecture to share with you in our upcoming Martech for 2027 report. (You can reserve first access to it here.)

In today’s newsletter, though, I want to dig into a time-scale view of context, the gaps AI is widening across those time scales, and how we might close them.

I find pace-layering diagrams to be an insightful way to understand complex systems. Not just computer systems. The idea was originally developed by architect Frank Duffy as “shearing layers,” and then expanded on by Stewart Brand to explain how different layers of a building change at different rates.

The site, structure, exterior skin, services (e.g., electrical, plumbing, HVAC), space plan, and stuff (e.g., furniture, appliances, art) all change on different timescales. Fast layers innovate, learn, and absorb shocks. Slow layers constrain, remember, and stabilize. The interesting challenges often aren’t within a layer, but at the seams where a fast layer rubs against a slow one.

10 years ago, in my book Hacking Marketing, I mapped marketing’s own pace layers, from slow (years) to fast (real-time): company, brand, campaign, channel, tactic, iteration, and feedback from individual customers. The dynamics between those layers are remarkably similar to those of a building.

Context — for both the company and the customer — is synthesized across all of these layers. A real-time activation for a customer happens in the context of a campaign or program, which happens in the context of the brand, which happens in the context of the business overall.

Not surprisingly, the bumps in marketing often happen as a result of friction or desyncs across these layers — when plays cause a campaign to drift out of alignment with other campaigns, programs, or the core messaging and values of the brand.

AI is now an earthquake, shaking marketing’s “building” across these layers — and opening up two significant gaps as a result.

While it’s always been challenging to keep fast-moving campaigns and plays aligned to the brand, the human-intensive processes and overhead required to launch them served as a natural brake — in a good way. While it constrained the number of things in flight, the time required to produce and approve them gave ample space for more strategic context to be considered along the way.

More people, with more experience, had more opportunities to flag when something felt off-brand, even if they couldn’t algorithmically articulate why.

But now? AI lets more marketers generate more campaigns — and more granular and specialized campaigns — more quickly than ever before. Almost every demo I see from a martech vendor these days shows how a few simple sentences typed or spoken to an agent can instantly spin up emails, landing pages, and follow-up, multi-channel nurturing sequences.

It’s impressive!

But this flood of creativity threatens to overflow the organizational levees that helped keep marketing activities in sync with the brand and the business. By expanding and accelerating execution, we’re also pulling strategic choices up into higher pace layers of marketing — yet maybe not consciously recognizing that’s what we’re doing.

Hey, greater speed and distributed creative energy can be powerful levers for competitive advantage. Or spectacular chaos if we let those forces run amok.

We can start by bringing more of our company context into our systems context, codifying more of our institutional knowledge in machine-readable form. That can help programmatically keep the whirlwind of AI creation aligned with the business.

But it’s not a once-and-done exercise. We also need mechanisms to keep that context up-to-date and have changes swiftly propagate across the myriad of campaigns, web pages, agents, sales enablement, email sequences, etc. that rely on it.

While we need software to wrangle the sheer scale of it all, software alone can’t solve it. We need to think carefully about where and when we have a human-in-the-loop — but also who that human is, what authority they have, and the time and tools we give them to make decisions.

Another gap is appearing higher up the pace-layer stack where we increasingly let AI optimize and adapt iterations and individual customer activations within a campaign. “AI decisioning” is the term most commonly associated with this.

LLMs let us personalize campaigns with more degrees of freedom, while reinforcement learning lets campaigns self-improve over time with a greater degree of autonomy. Together, they create the circumstances for campaigns to rapidly evolve in ways that aren’t fully visible to the humans overseeing them. There are just too many combinatorial possibilities.

If there are good guardrails in place — established to address the alignment gap we just discussed — this isn’t necessarily a problem. It’s a feature, not a bug. Evolutionary campaigns. (I still like the phrase “infinity campaigns” that Databricks coined earlier this year.)

The coordination gap isn’t between the campaign and its optimized iterations. It’s between the evolutionary adaptations of one campaign — or program or touchpoint — and the parallel AI-driven adaptations that are happening with the other campaigns, programs, and touchpoints across marketing’s universe.

If they evolve independently, rather than co-evolving, you can end up with a collection of individually optimized moments that are collectively fragmented in the eyes of the customer whose experience spans across them.

Like the alignment gap, the coordination gap has only recently opened up as a result of AI. The number of self-optimizing loops, the dimensions by which they’re able to evolve, and the speed by which those optimizations happen are all increasing.

Closing this gap will require technical innovation in “orchestration” across all these parallel agents. Yes, like context, orchestration’s a word that’s being thrown around a lot now, without a lot of details either. It needs to incorporate elements of control, coordination, cadence, and calibration wrapped around the context of the company and its customers.

We’re working on a framework that will cover all of that in Martech for 2027, so please do sign up for that release.

But three things are becoming very clear:

Context is where the frontier of martech innovation is right now.

Engineering that context is hard — which is why it’s a huge opportunity for martech platforms to deliver incredible value via context-as-a-service.

And finally: if context has been our word of the year for 2026, I think 2027 will be the year of “coherence” — how all these contexts, at all these layers, hold together as a whole. In other words, a brand.

Hopefully this was coherent.

Scott

P.S. WebEngage is one of our sponsors for Martech for 2027. Frans and I will be having a great chat with their CEO Avlesh Singh during our launch event on December 1. But in the meantime, you can check out what WebEngage is all about by clicking on the banner below.