GTM Engineering: How a Mid-Sized Company Replaced Its 40-Person Sales Team with an AI System

The story: 40 salespeople replaced by a system built on Clay

The company sold a no-code platform internationally, exclusively through email and LinkedIn, with no cold calling. Messages were personalized by hand and feedback was logged into the CRM manually, backed by a correspondingly large sales team.

Scalantec replaced that manual process with a system built on Clay and Lemlist, including automated CRM sync. For context on cost: with 40 salespeople and estimated fully loaded costs of around 100,000 euros per head, annual sales costs sat in the mid-single-digit millions, against tool and AI costs an order of magnitude lower, a factor Nicolas puts at roughly 100x.

The switch didn't happen overnight, it played out over roughly a year. The weakest performers were pulled out of the org first, some moved into other departments like marketing, not all of them let go.

The owner's decision came down to classic sales metrics: number of first meetings, reply rates, and closed deals. The system was later supplemented with cold calling, which Nicolas thinks can still work well depending on the industry, if it's done right.

What is GTM engineering?

The term go-to-market engineering was largely popularized by Clay, according to Nicolas, and describes a new role: using technology, automation, and tools to enable sales, build automated outreach, get clean data into the CRM, and prepare reps better for conversations. He says there are now hundreds of job postings with exactly this title.

The core toolkit is Clay, Claude Code, n8n, and the usual outreach tools (Lemlist, Instantly, HeyReach, EmailBison).

Step 1: Define the problem and the ICP

The starting point isn't volume, it's precision: finding the narrow set of companies for whom you solve the biggest problem. Nicolas recommends a closed-won analysis for this: enrich every past customer with revenue and firmographic data to find the 20 percent of customers driving 80 percent of revenue, a data-driven ICP instead of a gut-feel definition.

As an example, he cites Scalantec client Seven Senders, a logistics platform for cross-border parcel shipping in e-commerce. Their ICP: e-commerce companies with a certain shipment volume, identified via traffic data (e.g. Similarweb, Semrush) and a conversion rate formula. The resulting highly personalized campaign (referencing specific shipping countries and competitors) delivered, according to Nicolas, the following results:

More on the work with Seven Senders in the full case study.

Step 2: Build the total addressable market (TAM)

The ICP becomes a company list, the TAM. For more traditional industries, Nicolas recommends North Data, since databases like Apollo or Clay source their company data mainly from LinkedIn, where classic trade or construction-equipment businesses are often missing. For local businesses, he points to Google Maps scraping through the scraper marketplace Apify. As a current Apollo alternative with a larger database and verified email addresses, he mentions AI Ark.

Step 3: Find the right decision-makers

The typical targets are the economic buyer and the champion inside the target company. Where tools like Apollo, AI Ark, or Clay can't deliver contact data, AI agents step in, searching the company website or legal notice page for management names. From first name, last name, and domain, the email address can usually be derived reliably.

Step 4: Cold outreach that still works

Nicolas pushes back on the common assumption that more volume is the answer to falling reply rates. His starting point stays the precise list, paired with a message that triggers an emotional reaction, not the same sales structure over and over (trigger, problem, solution, social proof, call to action), which he says he can spot instantly as an AI-generated mass email.

Scalantec's approach: messages are drafted manually first, the way you'd actually write to that one specific person, then individual words or passages get AI-assisted personalization on top. Triggers like job changes, funding rounds, or tech-stack changes are mainly useful for timing, not phrasing, according to Nicolas, because lines like "Congrats on your funding round" are long since worn out.

For reference: average reply rates in cold outreach often sit at only around 2 percent according to Nicolas, and LinkedIn currently allows roughly 20 to 25 automated connection requests per day.

Why fully automated AI SDRs have failed so far

The conversation also covers 11x, a startup that promised a fully AI-driven SDR, including list building, research, and message sending. Nicolas's take: the hype was huge, the results weren't there, plenty of companies tried the tool, few stuck with it. His conclusion: people still buy from people, but subtasks like written outreach or clean CRM upkeep can, in his view, be fully replaced by AI.

What this means for salespeople

Asked what sales professionals should do given this shift, Nicolas is direct:

His recommendation: actively get hands-on with AI tools like Claude Code, n8n, or Sim AI, because your own output can multiply many times over as a result, in his view.

The future of GTM agencies

Asked whether automation eventually threatens the agency's own business model, Nicolas points to a thesis from the VC world that he cites in the conversation: for every dollar spent on software, roughly six dollars will flow into services, because generic software is too rigid and companies increasingly need individually tailored solutions. His expectation for Scalantec: services keep getting productized and automated, but customers will still need people for strategy, advice, and the customization built around AI agents, things like separate git repositories and Claude Code agents per client.