Real GTM engineering examples from 4 Belgian tech companies

Is GTM engineering hype or reality?

‍

The answer: it's definitely a reality, but not a new magical trend. It's a natural evolution of earlier marketing approaches, demand generation and growth hacking. What's new is who can build it: a small team with tools like Claude Code can now build the whole system itself, and these four companies did.

In practical terms, all four use it to find the accounts that are ready to buy now and reach the right person with the right message. It does the research and matching a seller would otherwise do by hand, so sellers spend their time on conversations.

All four companies weighed in on most of the questions asked during the event. For each question below, we took the company with the strongest takeaway.

1. How do you know which channels actually bring in customers? Spott, an AI-native workspace for recruitment agencies. Speakers: Kevin Vandeputte and Nathan De Troyer.
2. Should you hire a GTM engineering agency or build it in-house? Oper Credits, an AI agent for banks' credit processing. Speaker: Christine Hick, Head of Growth.
3. Which tools should you pick? Powernaut, an energy operations layer for retailers and independent power producers. Speaker: Roger Akayesi, Growth Marketing Manager.
4. Where does GTM engineering sit, and how do you stop builds dying? TechWolf, a skills platform for the Fortune 500. Speaker: Amaury Hellebuyck, revenue operations.

‍

‍

How do you know which channels actually bring in customers?

‍

Turn every channel into a signal, match each signal to one company, and weight the signals by how much they contribute to conversion.

Kevin Vandeputte and Nathan De Troyer went first, with how Spott runs go-to-market. Spott sells an AI-native workspace to recruitment agencies. In the ten months since their product became sellable, they went from zero to roughly $5M ARR and 400 agencies onboarded, on a five-person go-to-market team, four of them engineers by background.

They built their own system because a CRM can't answer two questions: which channels actually contribute, and who to call today. Organic LinkedIn, for instance, carries no trackable link, but it's one of Spott's biggest converters. Direct attribution scores it zero.

Spott reduces every channel to a signal: was this company served an ad, did they click it, which pages did they visit, how did they engage with a post. Everything goes into a staging table, then pipelines clean it up before the genuinely hard part: entity resolution. LinkedIn hands you an opaque identifier. Website tools hand you a company name with no domain, and there are a lot of identically named recruiting firms in the world. Spott matches on domains and LinkedIn IDs (never URL slugs, which change), falls back to a confidence score for anything softer, and sends anything below the threshold to a human. An intern did that work early on; the human-labeled set is meant to train an automated matcher they haven't built yet.

Once a company is resolved, every signal carries a weight estimating its contribution to conversion, so Spott can split pipeline value across channels even where nothing tracks the click directly. The weights start as guesses, so some are wrong at first. The team spots them because they know each channel well enough to see when a number doesn't add up.

What broke first: they built the front end before they'd figured out the data model, and next time they'd do it the other way around. Spott's advice for starting from zero is to set up the data model before connecting anything to it, and to build the do-not-contact list before the first campaign.

‍

‍

Should you hire a GTM engineering agency or build it in-house?

‍

Build it in-house when your market is small and timing matters, because an agency's volume playbook can't catch the right moment.

Christine Hick, Head of Growth at Oper Credits, sells an AI agent to banks, replacing parts of their credit-processing workflow. Her market is about 550 companies in Europe (for mortgages only). "We don't have to look for our customers," she said. "We exactly know who those are." That makes it a timing problem, not a volume problem.

She first hired a GTM engineering agency to build the outreach system, and they ran the standard playbook: Clay for enrichment, sequenced outbound messages, a cadence. As far as she can tell, it did not produce a single lead.

She doesn't blame the agency. Volume mechanics can't substitute for knowing the customer in a market this small, and what she actually needed was high personalization tied to real signal timing: knowing the moment a target contact was hiring a credit analyst, or talking publicly about AI. She had the idea by December. The agency, working inside Clay and email tooling, couldn't build it.

In January and February, Claude Code made it possible for her to build the system herself. She later added Hermes, an open-source agent, once she found Claude Code too workflow-bound for the LinkedIn signal work she needed. She built it herself and no longer relies on an outside team, though she'd welcome a dedicated technical hire: embedding this in CRM and workflow on sensitive banking data isn't something to cut corners on.

‍

‍

‍

Which tools should you pick?

‍

Pick tools for how well they connect to everything else, not for their feature list. A good API (a way for other software to read and write the tool's data) comes first, and an MCP connector (which lets AI tools like Claude use the tool directly) is even better.

Roger Akayesi is Growth Marketing Manager at Powernaut, which sells an energy operations layer to retailers and independent power producers. Its market is about 1,000 known companies, each with a full buying committee rather than a single decision-maker.

Powernaut's first setup, built with a GTM engineer, ran heavy automation into HubSpot and left sales to cold-call whatever landed. It wasn't working how they wanted, so they scrapped it and moved everything to Claude, building in-house.

The new pipeline runs on roughly 30 signal types: a job change into a relevant title, a job posting, a tender live or years old, a press release. When a signal matches, Claude scores it, enriches it through Clay and then Apollo to find the actual buying committee (not just any contact), and drafts a message tailored to each person's role. A salesperson only checks and sends, and deal stages move automatically on a response. The system also pulls from live conversations: something a rep heard on a call yesterday can seed tomorrow's campaign for a similar company.

Claude covered every step except outbound, because there was no connector for Lemlist, their outreach tool. Two months before the session, an MCP connector for Lemlist finally landed, and the whole flow closed. Roger's rule: pick MCP before API.

‍

‍

Where does GTM engineering sit, and how do you stop builds dying?

‍

Put the role centrally, and don't start a build until the team that will use it has named an owner.

Amaury Hellebuyck builds revenue operations at TechWolf, a Series B company selling to the Fortune 500. He also hosted the evening as an operator at Syndicate One. When TechWolf's go-to-market team started scaling from 30 people toward 60, its systems and processes hadn't kept up, so over the summer the company moved an existing engineer into a dedicated GTM-engineer role.

They debated where the role should sit: inside sales, inside marketing, or centrally. They chose centrally, for three reasons. What gets built (their signals project, for instance) serves sales, marketing and CS all at once. They didn't want to hire three separate people to solve the same problem three times. And they believe in centralizing operations rather than duplicating the same build across teams. The long-term target is three or four GTM engineers moving between team priorities from the centre.

Then they named the actual risk in their own model, in the most transferable line of the evening: build it, throw it into the function, and it dies. Their fix is a hard gate. No build starts until the receiving function names an owner who'll take it over once it ships.

Their prioritization filter is just as simple: every request either creates more opportunities for sellers, or frees up their time to work the opportunities they already have. Compress a day of outbound research into an hour, and a seller has nine more hours for actual customer conversations. No scoring model, just a conversation with the team.

‍

‍

What broke first?

‍

Every company hit a wall before it got this working, and the fixes are specific.

1. Spott: it built the front end before the data model, and its signal weights started as wrong guesses. Fix: data model first, and learn each channel before weighting it.
2. Oper Credits: an agency set up tools and a sequence that, as far as Christine can tell, produced no leads. Fix: she built it herself around signal timing with Claude Code.
3. TechWolf: a build thrown into a function with no owner dies. Fix: no named owner, no build.
4. Powernaut: a first setup with heavy automation into HubSpot left sales cold-calling, and Lemlist had no connector. Fix: they scrapped it, rebuilt in Claude and waited for an MCP connector.

‍

‍

Frequently asked questions

‍

What does a GTM engineer do day to day?

At Spott, Nathan describes it as closer to data engineering than software engineering: working out which data points exist, how they relate and how to turn them into a result sales can use.

Is GTM engineering the same as RevOps?

Amaury does both at TechWolf, revenue operations and GTM engineering. He describes the engineer as functional but technical: someone who can build, drive adoption and enable people.

When should a startup hire its first GTM engineer?

Amaury says it depends on stage. Early on, someone with Claude Code affinity can ship quick wins. The harder question, who maintains and extends what's built, arrives later.

How do you build a GTM engineering function from scratch?

Spott's advice: start with one channel, and get the data model right before connecting more. Understand one channel fully before moving to the next.

How many signals should a startup track for GTM engineering?

Powernaut runs roughly 30 types, including job changes into relevant titles, job postings, tenders and press releases.

What is entity resolution in a CRM, and why does it matter for GTM engineering?

It's matching an anonymous LinkedIn ID or a bare company name to the right company record. Spott calls it a very hard problem, because a wrong match feeds bad data into everything downstream.

‍

The three panelists agreed on one thing without being asked: you don't wait for the hire to start building.

‍

Syndicate One Field Notes is a portfolio knowledge-sharing series, run internally, for portfolio companies, by portfolio companies, so everyone moves faster. This session was open to the wider ecosystem, with speakers from two portfolio companies and two from the wider Syndicate One network.

Hosted at Beci, with an opening by Alain Heureux. Organised by Syndicate One, with Amaury Hellebuyck. Moderated by Julius Schelstraete, TechWolf. Presented by Nathan De Troyer and Kevin Vandeputte, Spott. Photos: Adrian Veithen.

Written by Ruta Fridrihsone, Syndicate One member and B2B marketeer.