Tom Husson: SaaS Is Dead, So Build Go-to-Market as a Relationship Engine Run by Agents

Tom Husson leads go-to-market engineering at Conveo, a team he describes as "basically just me and my agents." His argument for an AI-native GTM engine starts from a blunt premise: SaaS is dead. Software used to sell workflows, and go-to-market teams spent their days operating those workflows across a stack of tools. Agents now deliver the outcome, so what a company needs is a coordination layer that encodes how it works and remembers every relationship it has.

Husson made the case in conversation with Conveo co-founder and CEO Dieter De Mesmaeker on Conveo's podcast, recorded live at Cannes Lions 2026. The conversation covers what replaces SaaS, how to get a skeptical team to work alongside agents, and why Husson thinks the company that wins with AI is the one that maximizes its human potential.

Agents deliver outcomes, so the software layer gets thin

Husson picked up the idea that SaaS is dead, he says, from an AI lab he worked at before Conveo. The logic runs like this. A SaaS product sells a workflow. A GTM person then has to work that workflow by hand, wiring HubSpot to Marketo to the rest of the stack so each tool talks to the next.

Agents change the unit of work. You stop operating workflows and start assigning outcomes. Some coordination is still needed, and that used to live in standard operating procedures: here is how we reach out to these people, here is the sequence for this type of account. Husson's point is that you can now embed those procedures directly in the software the company runs on. At Conveo that layer is an internal app the team calls SuperG, which Husson describes as a coordination mechanism for agents.

He thinks about its purpose from first principles.

"What does the GTM function need to do? It's essentially build a relationship with your entire TAM."

Tom Husson, GTM lead, Conveo

He traces how that relationship has been faked for years. First came mass email lists, then first-name personalization, then openers that reference a prospect's LinkedIn activity, each automated until buyers saw through them. It is a cat and mouse game, he says. Real relationships are built on shared memories and many touch points, and a GTM engine should remember all of them, from the moment someone is a prospect to the moment they are telling everyone they know about the product. In his words, the question was "how can we essentially build a relationship building engine where it feels like you're building that one-on-one relationship with [Conveo]."

Building beats buying when adoption is the real problem

De Mesmaeker pressed on the hard part: building an app is one thing, getting people to use and trust it is another. Husson's first answer is that building in-house makes adoption easier. "HubSpot has to build for hundreds of thousands of customers so they have to make trade-offs that don't always work for us," he says. An internal team can build exactly what each person needs, and if one high performer works differently from the rest, the tool can bend to fit them.

His second answer comes from building four companies before this one, most of them around getting people to adopt new technology.

"It's always the people first, and so I made it easy to give me feedback, maybe a little too easy to give me feedback."

Tom Husson, GTM lead, Conveo

He keeps probing for friction, down to a button that does not open as fast as someone expects. Then he makes the agents fun. Each has a distinct personality, and he set them up to DM colleagues with messages each person was likely to answer. One colleague started out refusing to use the agents at all and, after a few of those exchanges, went through what Husson calls "an insane arc." His summary: "I genuinely think that making stuff fun is underrated."

The third move is finding champions. Husson points to a seller leading the effort who used to spend about two hours a day in prospecting workflows: building lists, filtering them, loading them into tools and coordinating LinkedIn outreach. That work is now automated and coordinated across the company. Website visitors from a given vertical get routed to the right seller, and a "next action" column tells the seller what to do. For a salesperson, Husson says, "the only thing that matters is how many meetings do I have of people that are ready to buy in my calendar."

The value compounds past sales. Conveo's research impact leads, its version of customer success, inherit every touch point an account has already had: what excited them on a call, what resonated and what did not.

The AI-native company wins by maximizing human potential

Husson calls GTM engineering a role that did not exist a decade ago. "It's company building through engineering, essentially," he says. Before, ops teams wrote SOPs and outsourced tooling that cost a fortune and rarely worked. Now software is close to instant, so a company can build its own way of working into its tools and keep control over things like messaging and sequences, which protects the brand.

When he hires, he looks for architectural intuition, ideally from a computer science background, plus a deep understanding of what makes a business scale and where its bottlenecks are. Asked which side will matter more in a few years, he bets on business understanding. He borrows a frame from AI safety, the trade-off between controlling models and aligning them, and draws the lesson that "you need to ride the exponential." Trying to hold it back is the losing move, and models will take care of more of the engineering.

That leads to the manifesto he has been writing. "The AI native company that wins is the one that maximizes its human potential," he says. The first implication is accountability: "you can't hold agents accountable." People carry responsibility for the choices their agents make, so the company needs high-agency, high-accountability hires who understand from first principles what those agents do.

"You need system builders."

Tom Husson, GTM lead, Conveo

One-off operators are exposed, he says: "they're going to get bulldozed by these agents," because a single person can build the agent that does their work. He sees some of that tension inside Conveo and coaches people to move to the side that directs the agents. The north star he set in his first week is to "help everybody operate at the speed of thought." He also notes that the strongest people use agents as coaches, so they keep getting better themselves.

The examples he and De Mesmaeker trade go well beyond sales: dashboards and a data pipeline built in-house instead of a BI tool, and an onboarding system built and personalized to the company instead of a bought learning management system. It echoes what engineering teams elsewhere describe in AI-native product development.

What it means for insights and marketing leaders

Husson's engine is a go-to-market system, but the logic transfers to anyone who owns customer understanding.

• Memory is the asset. The GTM engine is valuable because it remembers every touch point and hands it to the next person. Insights teams face the same problem with studies that end up in slide decks nobody opens again. Treating findings as reusable infrastructure is the research version of the same idea.

• Coordination beats more tools. If agents produce outputs, the scarce thing is a thin layer that says how your organization works and routes the right information to the right decision. For research, that means a voice of the customer program wired into the moments where teams decide, so evidence reaches the decision while it is still open.

• Adoption follows the person. Husson's three moves, easy feedback, a bit of fun and visible champions, apply to any new research workflow a team is asked to trust.

• Accountability stays human. Agents can draft, route and summarize. A named person still owns the decision and the interpretation, which is exactly where insights professionals add the most value.

A relationship engine needs a clear picture of the people on the other side

Husson closes by naming the conditions that let Conveo build this way: real pressure from fast growth, which makes people want new tools; funding to fuel the work; and a team small enough to put something in production the day after it is built. Big companies lack that pressure, he says, so even a strong new tool ends up pushed onto old systems.

His governing idea holds for any function: agents take on the workflows, and people stay responsible for the relationships and the judgment. The same applies on the research side. Conveo's AI research assistant runs video conversations with real consumers, so insights teams can get to the why behind behavior faster, and the people who know the business decide what to do with it.