There’s a sequence in Star Wars’ Andor where the pilgrims walk to the Eye of Aldhani. Once a decade, a meteor storm lights up the sky above a remote plateau. The faithful trek for days, sleep on rocks, and look up. It’s beautiful. It’s also, in the show, the perfect cover for a heist. Everyone’s distracted by the lights.

I keep thinking about that scene as May 5 approaches.

That’s when Scott Brinker and Frans Riemersma debut the State of Martech 2026 report and the updated 2026 Marketing Technology Landscape on MartechDay. It’s the closest thing our industry has to a recurring religious event. Tens of thousands of marketers, vendors, consultants, and operators all turn their faces toward the same supercut of category logos and survey data. There’s a webinar, a keynote, a flurry of LinkedIn posts dissecting the charts and tables. I’ll be there too. I’m not above the pilgrimage.

The interesting question isn’t what the lights show. It’s what happens in the quiet while everyone’s looking up.

What the December warm-up already told us

The Martech for 2026 report dropped with some fanfare back in December, and most of the industry seems to have moved on already. That report is the warm-up act for what’s coming on May 5, and the numbers in it are doing real work.

The number I keep returning to is 90%. Ninety percent of marketing organisations say they use AI agents in some capacity. Sounds enormous. Then you read the next line → only 23.3% have any agent in full production, and the rest are piloting, experimenting, or running them in narrow workflows with a human approving every output. McKinsey’s number for the same population is closer to 1% fully scaled, which tells you how loose the term has become. If your “AI agent” is a chatbot answering FAQs, you can claim production deployment with a straight face.

Brinker and Riemersma also introduced a frame I’ve used with three clients already. They split the marketing operation into a Factory and a Laboratory. The Factory protects current revenue. Deterministic, governed, boring in the way that production systems should be. The Laboratory is where you experiment with agents, test new approaches, accept that things will break. Different KPIs. Different tolerance for failure. Different people, too, probably.

Most organisations are trying to run both with one team, one set of metrics, and one governance model. That doesn’t work. I’ve watched it not work in those three client engagements this year alone.

What I’m actually watching for on May 5

So what do I want the State of Martech 2026 report to tell me, beyond the headline meteor storm?

First, I want to see what’s happened to Brinker’s composable canvas idea since the Databricks-sponsored paper landed in March. He argued that the rigid stack is dissolving into a fluid canvas, with five classes of data (customer, company, content, code, control) sharing a unified foundation. It’s a thoughtful frame and a serious piece of work. It’s also, as David Chan pointed out in his response on Medium, a frame that emphasises the technical shift (into an even higher gear) more than the organisational one. Decision paralysis, governance, role ambiguity. The canvas is composable. The organization chart isn’t.

I want the May report to tell me whether companies are actually moving toward this canvas, or whether they’re nodding at it in slide decks and going back to arguing with their CDP vendor about identity stitching the next morning. My suspicion is the latter, and I’d love to be wrong.

Second, I want vertical breakdowns that go beyond the usual aggregate. Last year’s report leaned on overall percentages, which flatters everyone and informs no one. The interesting question is which industries are pulling ahead and why. I’d bet financial services and retail are further along, partly because they’ve had decisioning engines for two decades, partly because regulators forced them to invest in data lineage long before AI made it fashionable. I’d bet manufacturing, public sector, and parts of healthcare are still figuring out what their Martech stack contains. The gap between those camps is widening, not closing, and I want the data to show it. I’ve written before about same stack, different worlds, and I suspect 2026 will make that pattern impossible to ignore.

Third, and this is the one that matters most to the people I work with, I want to see what’s happening to roles.

The middle is dissolving

Here’s where the AI conversation gets uncomfortable, and most reports skirt it. The middle layer of the marketing operations organization is dissolving. Not the senior strategist roles, which still need someone who can argue with the CFO. Not the genuinely technical roles, the engineers and the architects. The middle. The campaign manager who pulls a list, the analyst who builds the same weekly dashboard, the automation specialist who configures the email send. An agent can do a passable version of all three before lunch.

I’m not saying that’s good. I’m saying it’s happening. And the people sitting in those seats often don’t know it yet, because their managers haven’t figured out how to tell them.

What’s emerging in its place is messier than the LinkedIn thought leaders make it sound. AgentOps is the term Joao Moura at CrewAI keeps using, the discipline of running fleets of agents the way DevOps runs fleets of services. Some companies are calling it Agent Factory. Rebecca Corliss at GrowthLoop frames it as marketing ops evolving into AI orchestration. Whatever the title ends up being, the work is the same. Designing agent workflows, monitoring their cost and reliability, catching the 20% that’s wrong before it goes out the door. I’ve sketched my own version of this in the new AI org chart, and the shape of it keeps shifting every quarter.

That last bit is the actual skill. Look, I’ll be honest, the marketers who’ll thrive in 2026 aren’t the ones who can write the prompt. The ones who’ll thrive are the ones who can read the agent’s output and instantly know which fifth of it is hallucinated, why, and what to do about it.

That’s domain knowledge. An engineer can build the agent, monitor its uptime, manage its cost. An engineer cannot tell you that a 4% open rate is fine for a B2B reactivation campaign and disastrous for a retail flash sale. That’s vertical-specific pattern recognition, layered with editorial judgement, built from years of seeing what good and bad actually look like in the work. Which is awkward, because the easiest way to build that knowledge was to spend a few years in the middle-layer roles that are now disappearing.

So we’re heading into a world where the entry-level rung of the ladder is being sawn off, and the senior rungs require a kind of expertise that can only be built by climbing the rung that no longer exists. I genuinely don’t know how that resolves, and I’m suspicious of anyone who says they do.

The maturity question, honestly

McKinsey’s October 2025 piece on rewiring martech put it plainly. Sixty-five percent of B2C organisations self-identify as developing or operational in martech maturity. When the researchers interviewed those same respondents, the proportion claiming operational maturity dropped significantly. People rate themselves higher than the evidence supports. This isn’t new. It’s been true for fifteen years.

What AI is doing is widening the gap between the organizations that built the muscle and the ones that bought the tools. Tools without muscle is the McKinsey finding restated. It’s also exactly what I’ve been arguing in the martech readiness gap for a while now. Structure, capability, and process matter more than the platform. The tooling has finally caught up to the point where bad organizations can no longer hide behind a Salesforce contract.

The leaders pulling away in 2026 are the ones who quietly invested in operational fluency while everyone else was buying licences. The laggards are still trying to AI their way out of an organization design problem.

What I’ll be looking for in the lights

When May 5 rolls around and the report drops, I’ll be reading it for three things. Whether the composable canvas is real adoption or aspirational PowerPoint. Whether the vertical gaps are widening the way I think they are. And whether anyone is finally publishing data on what’s happening to the middle of the marketing organization chart, because that’s the conversation we keep avoiding.

The Eye of Aldhani is gorgeous. The pilgrims come back changed. But the real story of 2026 is happening in the shadows while everyone’s looking up.

If you’re reading the report next week and not sure what it actually means for your stack, your team, or your roadmap, that’s exactly the kind of confusion the Martech Workshop is built to work through. Otherwise, drop me a note. I’d genuinely like to compare notes on what you spot.