AI in Sales: What Actually Works in GTM in 2026

The starting point: data chaos at a logistics platform

The client is a European B2B logistics platform for international e-commerce, with several hundred employees and several salespeople. The team believed it knew its market. But nobody had ever mapped that market systematically, and again and again a relevant company came up that nobody had ever been in touch with.

Then there was the CRM:

• contacts who had long left the company
• wrong email addresses and missing phone numbers
• accounts and opportunities that salespeople created even though they already existed three times
• a website visitor tool whose data nobody looked at
• inbound requests that someone checked by hand
• touchpoints from the marketing channels that were tracked but never brought together

Nicolas calls the result data chaos and process chaos, and it showed in the pipeline and in conversion.

The approach: audit, foundation, activation, reporting

It started with a go-to-market audit: conversations with the Head of Marketing, the Head of Sales, RevOps and some of the salespeople about their goals, plus a close look at the CRM, its duplicates, pipelines, dashboards and reporting. From that came a strategy in three steps.

1. The foundation

• Capture the whole market. All companies from North Data, which lists every company in the German commercial register, plus databases such as Apollo and AI Ark.
• Enrich and prioritize. The data that matters for this client: which shop system, how much website traffic, which tech stack. From that, a segmentation and an ICP score.
• Find the decision makers. The relevant people at every company, with information that helps sales in the conversation, and with email addresses and phone numbers.
• Clean up the CRM. Remove duplicates, update outdated records and upload the companies and contacts nobody knew yet, so the CRM is the single source of truth again.
• Track signals. For the top accounts: job changes, promotions, changes on the website, and social signals such as likes, comments or new followers of the LinkedIn company page.

2. Activation

All of this data comes together in a lead score. Salespeople know which accounts to focus on this week: the ones that fit the ICP well and are active right now, for example because they visited the website or interacted on LinkedIn.

With a small market of maybe 2,000 companies, Nicolas advises against wasting chances on generic, automatically sent AI messages. Clay can write very well personalized emails, but the real challenge is the copywriting.

3. Reporting

In the end, all of the data has to show up in dashboards: how much activity is happening, how much outreach goes out, how many new deals come in.

CRM or Clay: where what happens

At most companies the CRM, whether Salesforce, HubSpot, Dynamics, Attio or Pipedrive, is the single source of truth. But every CRM has its limits, and that is what Clay is for. Nicolas describes Clay as a data orchestration tool: it brings together data from many databases and cleans and filters it. The market and contact lists in this project were built in Clay.

• A waterfall instead of either/or. Clay has more than 200 native integrations. You used to pick one provider for phone numbers and one for emails. In Clay you query several databases in turn and have the results validated. If a record isn't found or is invalid, Clay moves on to the next source. That gives you data for far more contacts.
• AI agents for research. Salespeople and RevOps managers build AI agents in Clay, for example to pull the managing director from a company's legal notice. For 50 or 100 companies, ChatGPT or Claude can do that too. With 10,000 contacts, each needing their LinkedIn profile and their posts from the last seven days, those tools reach their limits or hallucinate.

How long a project like this takes

The market analysis alone, meaning finding, scoring and enriching the companies, takes 4 to 8 weeks for smaller markets according to Nicolas, and can take several months depending on how much data is needed. Everything together, including signals and tasks for sales in the CRM, takes 6 to 9 months on average. When it has to go faster, Scalantec puts more GTM and RevOps engineers on the project.

Where companies should start

Many go-to-market teams lack a good strategy, Nicolas says. They try channels and tactics they picked up on LinkedIn: a few thousand euros a month on Google Ads, or messaging everyone on LinkedIn. Without a system, without infrastructure and without a strategy. What has to be clear first: which channels you really want to serve, which tools make sense in the setup, and how the organization needs to be built.

Host Dominic Blank sums it up: AI doesn't fix your process. It helps where you already know what works and what should be automated.

What surprises clients today

From his exchange with American and British agencies, Nicolas sees Germany, and the Mittelstand in particular, still lagging behind in parts. Even automating LinkedIn and email outreach or account and contact research often gets an amazed reaction there. B2B SaaS companies and scale-ups in Berlin and Munich have worked with tools like these for longer.

What almost always surprises: the market mapping. Companies that believe they know their market suddenly have 20% more companies on their radar, and clean data for once.

Sales cockpits

New and in high demand right now are sales cockpits: a kind of operating system for sales where everything comes together. It can live in the CRM, but many companies want a separate tool for it. At the Clay community's two-year anniversary, the founding GTM engineer of Peec AI showed his team's cockpit: a web app where the salespeople see all signals and companies, who is already in the CRM, who has been in touch, website intent, social engagement, the last conversation and the activity history.

Three quick wins

1. Go through your own LinkedIn connections. Download your LinkedIn connections and check with Clay (or with Codex) who fits your ICP and has the right title. You are already connected with these people: the message arrives, it doesn't land in spam, and they see your profile right away. This works if your buyers are active on LinkedIn.
2. Work through closed-lost deals systematically. You have already been in touch, and they know your name.
3. Keep at it. Sales is often a long game across several channels. AI, automation and tools help and show the direction, but they don't replace the personal conversation.

Nicolas also recommends staying curious: there is a lot of good content from practitioners on YouTube and LinkedIn. If he had to pick one person to follow, it would be Jordan Crawford, a GTM expert from San Francisco whom Nicolas has met several times in Berlin.