What I actually took away from The MarTech Summit Madrid
- 6 hours ago
- 9 min read
By Andrew Poole - Specialist Marketing Consultant - Sojourn Solutions
Before attending The MarTech Summit Madrid, I wrote about what I was hoping to take away from the event.
I wanted practical conversations about Marketing Operations, AI governance, customer data, sales and marketing alignment, and the reality of making MarTech work inside complex organisations.
Having now attended, the short version is this: The event itself was excellent. The organisation was slick, the venue was fantastic, the agenda was relevant, and the quality of the non-vendor speakers was genuinely strong.
But the biggest takeaway for me was not what I expected.
It was not that AI has come for Marketing Operations.
It is that AI is being heavily pushed into Marketing Operations teams by companies that, in some cases, do not seem to fully understand the operational environment they are entering. And that should make marketing leaders pause - then panic.
First, credit where it’s due
The MarTech Summit Madrid was very well put together.
The location was brilliant, right in the heart of Madrid, and the venue at VP Plaza España Design worked extremely well for the format. The room setup made sense, with the main stage close to the vendor area, enough space to move around, and a flow that made the day easy to navigate.
The agenda itself covered a strong mix of topics, from customer experience and digital experience through to AI, accountability, social strategy, revenue alignment, brand, and the evolution of the marketing function. The official session slides were shared after the event under the theme “Redefining MarTech: Autonomous, Intelligent, Human-Centred”, which is a fairly accurate summary of where the day’s conversation seemed to sit.
The non-vendor speakers were knowledgeable, credible, and from a good mix of organisations. That mattered. There was enough breadth in the sessions to reflect the day-to-day reality most marketing teams are dealing with: Data, customer journeys, automation, alignment, brand consistency, AI adoption, team structures, and the ongoing battle to make technology actually do what the brochure said it would do.
So no complaints there. As an event, it delivered.
As a snapshot of where MarTech is heading, it was even more interesting.
And, in places, more worrying.
The AI agent gold rush has arrived
The most noticeable thing on the vendor side was not the variety of tools.
There was variety, at least on paper. Vendors covered everything from mobile security and web page creation to automation platforms, brand systems, font management, and customer engagement technology.
But scratch the surface and a lot of them were really there to talk about the same thing. AI agents.
Not always in exactly the same language, but close enough. AI agents that could build campaigns. AI agents that could generate content. AI agents that could orchestrate end-to-end marketing activity. AI agents that could help marketing teams move faster, do more, reduce effort, and automate huge chunks of the process.
At times, it felt as though you could have swapped the logos on a few of the decks and the message would barely have changed.
Everyone had arrived at the same party wearing the same outfit, then acted surprised when the room looked familiar.
That does not mean the technology was bad. Some of it was genuinely impressive. There were interesting ideas, useful capabilities, and plenty of examples of how AI could remove friction from marketing workflows.
But the repetition was hard to ignore.
It felt like a lot of companies had looked at their existing product set, looked at the AI agent conversation happening around them, and decided they needed a version of it immediately.
Not necessarily because the market had fully worked out the operational problem.
More because nobody wants to be the vendor left saying, “we do not have an AI agent yet.”
The missing conversation: Governance
This is where the day became most interesting from a Marketing Operations point of view.
There was a lot of talk about what AI agents could do. There was far less discussion about what they should be allowed to do.
That distinction matters.
Because an AI agent that can create a campaign is not just a content tool. It is potentially interacting with data, systems, approvals, audiences, assets, customer segments, consent rules, sales processes, reporting structures, and brand controls.
In an enterprise environment, campaigns do not simply appear because someone in marketing had a nice idea and a prompt window. They involve multiple teams.
Marketing. Sales. Legal. IT. Data. Security. Brand. Compliance. Regional teams. Sometimes procurement. Sometimes finance. Sometimes the person who knows why one specific field in the CRM absolutely must not be touched because of something that happened in 2019. That is the reality.
And yet, too often, the AI agent story seemed to glide straight past it. The narrative was often: Prompt goes in, campaign comes out. Lovely.
But where is the approval process?
Where is the legal review?
Where is the data access control?
Where is the PII risk assessment?
Where is the audit trail?
Where is the disaster recovery plan?
Where is the human-in-the-loop checkpoint?
Where is the rollback process if something goes wrong?
Where is the definition of what the agent can change, what it can recommend, and what it must never touch?
Those questions are not boring admin. They are the difference between useful AI adoption and a very expensive incident report.
Everyone can make the sausage now
One line from the State of MarTech 2026 stuck with me:
“Everyone’s learned to make the sausage with AI. Almost nobody’s bought a labelling machine.”
That was exactly the feeling.
A lot of the market has learned how to produce AI-powered outputs. Content, campaigns, workflows, recommendations, journeys, summaries, segments, dashboards.
The sausage machine is running.
But the labelling machine is governance.
It tells you what happened. Who approved it. What data was used. What system was touched. What risk was introduced. What decision was automated. What should be reviewed. What should be blocked. What can be trusted.
And right now, that part feels dangerously underdeveloped.
This is especially true when AI agents move from “help me write a subject line” to “help me build and activate this campaign.” Those are not the same thing.
One is assistance. The other is operational control. And the closer AI gets to operational control, the more important governance becomes.
Dark AI is already here
Another issue that did not get enough attention is the reality of dark AI use.
By that, I mean the AI activity already happening inside organisations without formal approval, visibility, documentation, or control.
People are using AI tools whether businesses have a strategy or not. They are using them to summarise customer information, draft emails, rewrite content, analyse spreadsheets, create campaign logic, build presentations, generate code, and speed up all the awkward little tasks that sit between strategy and execution.
Some of that usage will be harmless. Some of it absolutely will not be.
The risk is not just that someone uses AI badly. It is that the organisation has no idea where AI is being used, what data is being pasted into it, what outputs are being trusted, or how those outputs are making their way into customer-facing activity.
That is before you even introduce vendor-built AI agents that connect more deeply into marketing systems.
So when AI agent adoption is discussed as though the main challenge is excitement, speed, or productivity, it misses the bigger operational issue.
Many companies are not starting from a clean, governed AI environment. They are starting from hidden usage, unclear ownership, inconsistent policies, and a growing pressure to “do something with AI.” and that is not a foundation, that is a wobbling table with a very expensive vase on top.
The networking felt different too
A smaller observation, but still worth mentioning: Networking felt harder than it used to.
That is not a criticism of the organisers. The event setup gave people plenty of opportunity to speak to each other.
But the behaviour in the room felt different.
A lot of companies seemed to have sent pairs or small teams. People were friendly, but many naturally stayed close to the colleagues they came with. It felt, in some cases, like attendees were enjoying a rare chance to spend time with their own teams away from the office or home-working routine. Completely understandable. But it does change the dynamic.
By the later sessions, a noticeable number of people had already left, and the post-event drinks felt heavily vendor-led. That made the networking less open than I remember from pre-Covid events, where people often seemed more prepared to walk up to strangers and start a conversation without first performing a full emotional risk assessment.
Again, not a complaint. More an observation.
The event was strong. The people were lovely. But the shape of networking has changed.
The real takeaway: AI agents expose the operating model
The conclusion I came away with is not that AI agents are bad. Far from it.
AI agents have a very real role to play in Marketing Operations. They can support campaign QA, workflow creation, content checks, data review, journey planning, reporting, documentation, governance monitoring, and plenty of other tasks that currently consume too much human time.
Used well, they could be genuinely transformative. But the phrase doing a lot of work there is "used well."
Because the problem is not whether companies can build AI agents. Clearly, they can.
The problem is whether they understand the Marketing Operations environment those agents need to work inside.
That environment is messy. Cross-functional. Political. Regulated. Data-heavy. Process-heavy. System-dependent. Full of edge cases, exceptions, legacy decisions, regional differences, and commercial pressure.
An AI agent that ignores that complexity is not a solution. It is a demo. And demos are easy.
Operational maturity is not.
This is where Marketing Operations matters more than ever
For me, the event reinforced something I already believed, but with a bit more force.
AI adoption in marketing is not mainly a technology challenge. It is a Marketing Operations challenge.
The businesses that get this right will not be the ones that simply buy or build the flashiest AI agent. They will be the ones that understand how to introduce AI into the real machinery of marketing without breaking trust, compliance, data integrity, customer experience, team confidence, or brand reputation.
That means asking harder questions upfront:
What should AI be allowed to do?
Which systems can it access?
Which data is off limits?
Where does human approval sit?
How are decisions logged?
What happens when something fails?
Who owns the agent once it is live?
How is ROI measured?
How does legal get visibility?
How does IT stay comfortable?
How does sales trust the outputs?
How does marketing avoid creating a faster version of the same broken process?
These are not questions to answer after implementation. They are the implementation.
Why experience still matters
This is why breadth of Marketing Operations experience matters so much.
It is easy to talk about AI agents in isolation. It is much harder to understand how they fit across marketing automation, CRM, customer data, campaign operations, consent, sales handoff, reporting, governance, and organisational process.
That is the bit companies cannot afford to skip - yet seem to be doing so...
Because introducing AI agents into Marketing Operations is not just about building something clever.
It is about building something safe, useful, measurable, and commercially defensible.
It needs governance from the start.
It needs a human-in-the-loop approach.
It needs clear controls around what AI can touch, what it can recommend, and what it can activate.
It needs involvement from the teams that will be affected, not just the team that got excited in the vendor demo.
And, yes, it needs enough red tape to stop the business from accidentally burning through budget, creating legal risk, exposing customer data, damaging the brand, or giving the CFO yet another reason to ask why marketing bought a thing.
I’m not being dramatic - the risk deserves a bit of theatre.
Final thought
The MarTech Summit Madrid was a strong event. Well organised, well located, and full of relevant conversations.
But the thing I will remember most is not a single session or slide. It is the wider signal from the room.
The AI agent race is on.
Vendors are moving quickly. Marketing teams are under pressure to keep up. The technology is becoming more capable. The promises are getting bigger.
But governance, operational reality, and cross-functional ownership are still lagging behind... massively!
That gap is where the danger sits.
It is also where the opportunity sits.
Because companies that take AI agents seriously, not as a novelty but as part of their Marketing Operations infrastructure, will have a huge advantage. They will move faster, yes. But more importantly, they will move safely, with control, visibility, and a clearer link to business value.
That is the difference between adopting AI and simply adding to the chaos.
And after Madrid, that difference feels more important than ever.
Andrew Poole is a Specialist Marketing Consultant at Sojourn Solutions, where he spends much of his time thinking about Marketing Operations, MarTech, AI governance, and why perfectly good marketing teams are still being held hostage by broken processes and suspicious spreadsheets.
