I want to be precise about what this piece is and isn't. It isn't a framework for evaluating founders in general — there's no shortage of those. It's a description of specific signals we've learned to look for over five years of seed-stage investing in AI marketing technology, drawn from diligence conversations with founders across Europe and beyond. Some of these signals are counterintuitive. Some contradict what you'd find in generic founder evaluation rubrics. All of them are real patterns we observed before we started articulating them.
A note on the category: AI MarTech has specific dynamics that differ from, say, AI infrastructure or AI coding tools. The buyers are marketing teams and CMOs — practitioners who have been burned by overpromised marketing technology many times before. The product has to demonstrate value in weeks, not quarters, because marketing teams don't have patience for long deployment cycles. The competitive landscape includes large incumbents with distribution advantages, horizontal AI platforms, and an endless supply of point solutions with similar feature sets. These dynamics shape what makes a founder distinctive in this category.
Signal 1: They describe a buyer's frustration, not a technology capability
The first question I ask every founder is some version of: "Why does this need to exist?" The answers fall into two categories. Category one: "Because we can now do X with language models that wasn't possible before." Category two: "Because marketing teams doing Y have this specific frustration that no existing tool solves well."
Both can be valid starting points. But in practice, the founders who lead with technology capability often build products in search of a buyer, while founders who lead with buyer frustration have already done most of the market validation before they write a line of code. The frustration-first founders can typically name the specific job title of the person with the problem, describe exactly what their day looks like when they encounter the problem, and tell you what they're currently doing as a workaround. That level of buyer specificity almost always correlates with faster initial sales cycles and lower churn, because the product was designed around a real workflow rather than a plausible one.
We've backed founders who can describe in precise detail what a VP of Content at a mid-size SaaS company does between 9am and 11am on a Tuesday — because they were that VP, or they spent three months sitting next to one before building anything. That's the kind of buyer specificity that makes a product feel inevitable rather than interesting.
Signal 2: They have a considered view on AI quality thresholds
In AI MarTech, the question "how good does the AI output need to be?" is not as simple as it sounds. A founder building for high-volume social media copy can accept lower quality thresholds than a founder building for regulated financial communications. The quality threshold determines the model architecture choice, the evaluation framework, the human-in-the-loop design, and ultimately the unit economics of the product.
Founders who haven't thought carefully about their quality threshold often have a dangerous assumption: that "good enough" quality will be obvious from user feedback. In practice, marketing content quality is subjective and context-dependent, which means quality problems often present as customer churn rather than clear bug reports. By the time you understand that your quality bar was too low for your specific use case, you've already damaged customer relationships that took months to build.
The founders who impress us can articulate their quality threshold precisely, explain why it's the right threshold for their specific use case, describe how they measure against it, and discuss what happens when outputs fall below it. That conversation usually takes about fifteen minutes and tells us a lot about how seriously they've thought about building a production system versus a demo.
Signal 3: They understand the procurement structure of their first buyer
This signal matters more in European enterprise markets than in US PLG-first markets, and it's something I focus on specifically in GTM diligence. Many founders building for marketing teams assume the CMO is the decision-maker. In European enterprise, the reality is often more complicated: IT has a seat in any procurement involving data, legal has a seat in any procurement involving AI-generated content, procurement itself has a seat in anything over a certain contract value, and the CMO is one stakeholder rather than the decision-maker.
Founders who haven't mapped this procurement structure build products that CMOs love and that die in procurement. The product doesn't have the security documentation for IT, doesn't have the compliance features for legal, doesn't have the contract structure that procurement can process. Every one of those blockers costs sales cycles that an early-stage company can't afford.
We look for founders who have either navigated this structure before (from their operating background) or who have done enough discovery with actual target buyers to understand it. The tell is whether they can describe the buying committee, not just the buying champion.
Signal 4: They have a specific answer to the cold-start problem
The cold-start problem in AI MarTech is real and frequently underestimated. A personalization system that needs behavioral data to personalize doesn't personalize well on day one. A brand voice model that needs to be calibrated on a customer's content corpus doesn't produce on-brand content before it's been calibrated. A content performance system that needs historical performance data to predict performance doesn't predict well for new customers.
Every product with a data flywheel has a cold-start problem. The founders who have good answers to it have usually thought carefully about the deployment experience — specifically, what does the product do during the period before it has enough data to deliver its differentiated value? This might be: a manually curated baseline capability that's genuinely useful while the model calibrates; a library of pre-built templates for the most common use cases; a configuration wizard that jumpstarts calibration through structured questions; or a different pricing model that manages customer expectations during the ramp-up period.
Founders who say "customers will see value once they've been using the system for a few months" are often right about the value and wrong about the patience. We push on this hard and look for founders who have already designed their way around it.
Signal 5: They have a theory of expansion, not just a theory of acquisition
At seed stage, most of the attention is on the initial sale — what's the wedge product, who's the first buyer, how does the first deal close. These are the right questions. But in AI MarTech specifically, the trajectory from first deal to sustainable business almost always runs through product expansion within existing accounts rather than pure new logo acquisition.
The reason is that the initial land is often small — a single use case, a single team, a limited data set. The economics that make the business work require expanding the contract, adding use cases, and increasing the data and workflow surface area within the account. Founders who have thought about this expansion path carefully know which product capabilities unlock the second and third use case, how the data from use case one makes use case two better, and which buyer relationships cultivated in the initial deal open doors to the larger enterprise-wide opportunity.
We're not saying initial acquisition doesn't matter — it obviously does, and without the first deal the expansion theory is irrelevant. The point is that in B2B AI MarTech, the first deal often looks small and the business case lives in the trajectory. Founders who can describe that trajectory in specific rather than generic terms — "after we prove value in the email channel, we expand to landing page personalization because the same audience model applies and the buyer's already in our product" — are building with the right mental model.
These five signals don't constitute a complete framework. There are founders we've backed who were weak on one of these and exceptional on others. There are founders who had all five signals and still struggled because the market timing was wrong or the team dynamics broke down. Seed-stage investing is not a checklist exercise. But these patterns are real, they are predictive in our experience, and they are the specific things we're listening for when we meet a MarTech founder for the first time.