There's no shortage of content about AI in GTM. Signals, orchestration, ABM, outbound, it's everywhere. What's harder to find is someone willing to open up their actual system and show you exactly how it works, what broke along the way, and what the numbers look like after six months of running it.

That's what we built Brains of GTM (our new webinar) for and our first session delivered exactly that.

Mark Kosoglow, CRO at Docebo, and Kevin Driscoll, Head of Global ABM at Datadog, came in with two very different builds and the same outcome: more pipeline, with the same team. Here's what they shared.

The problem neither of them could ignore

Before either of them got into the mechanics of what they built, they both put their finger on the same underlying problem.

Most revenue teams are running on disconnected signals. Your CRM holds some of it.  Your ad platform, your SEP, your enrichment tool, they each hold a piece of the picture. And every day, your reps are expected to mentally triangulate all of that, decide who to contact, figure out what to say, and execute at volume.

As Mark put it: "It's too many variables for the human mind to comprehend, weight, score, figure out, keep track of, be aware of, and factor in."

Kevin's version of the problem was framed around ABM specifically: most teams stop at account warming. Send some display ads, heat up the account, hand over to the SDRs. Done. But he argued that this is where ABM's job starts, not ends, and that the teams winning today are the ones pushing all the way through to person-level action.

Two different angles on the same gap: the missing intelligence layer between your systems of record and your systems of action.

What Mark built: the GTM brain that runs the rep's day

Mark's VP of Business Development has a useful way of describing the challenge. Every BDR makes five key decisions for every outreach:

1. Which account deserves my attention right now?
2. Which person inside that account should I contact?
3. What message will resonate with them?
4. What cadence should I follow?
5. What's the CTA that will make them respond?

If a BDR gets all five right, consistently, they perform. But most don't because making five high-quality decisions per prospect, hundreds of times a week, across dozens of signals, is genuinely impossible to do well manually.

So Mark's team automated most of them with UserGems.

Using UserGems' Data Agents and Intelligence Agents, they built a GTM brain that ingests signals from across their stack: job changes, website visits, event engagement, closed-lost history, ad engagement , combines them into a single weighted score, and each morning queues every BDR a prioritised list of who to contact, already loaded into the right sequence in Outreach, with a Gem-E-written personalised email ready for review.

The rule they operate by: a contact only enters a sequence when enough signals stack up to justify it. And when they do, the message only uses two of those signals, it's enough to be relevant, not so many that it reads like a surveillance report.

"We do not do a job change play. We do not do a website visitor play," Mark said. "We use those as signals and when working together, someone is more likely to be worth going after. If you see a job change and someone on your website, you can write a much more convincing message than if you just hammer them with one signal."

The results:

• Sub-2% reply rates → 11–16%, depending on the play
• 100+ enterprise opportunities generated in 6 months
• First opportunity landed within 4 days of turning the system on
• 30x jump in reply rates in European markets within 2 weeks of enabling localised sequences
• 56% of all BDR activity now driven by the brain, with 14,500+ contacts moving through it
• 80% of rep time now spent on the top 20% of highest-converting accounts, up from 60% of time wasted on non-selling tasks

"UserGems was a tool for us before we leveraged it in a strategic fashion," Mark said. "This is now much more than a tool. It is something that would be almost impossible for us to pull out."

What Kevin built: a surround-sound ABM engine that optimizes for meetings, not warmth

Kevin's philosophy starts with his key advice: stop measuring whether accounts are warm. Start measuring whether you're getting meetings.

"I don't think deals get done without meetings," he said. "Even meetings that are unproductive give you more information than you had before. If you get meetings, the pipe will take care of itself."

His team's answer to that is Pipeline Generation Days: structured sprints where SDRs are given pre-built contact lists, AI-generated messaging, and a single goal to book meetings. The results were extreme enough that RevOps had to revise the monthly quota targets mid-year because the team was blowing past them.

But PG Days are just one part of a broader surround-sound motion Kevin has built by himself. A few things that stand out:

They go where their ICP actually is. Kevin's buyers are technical , they're infrastructure engineers, platform teams. They don't spend their day on LinkedIn. So Datadog's ABM motion now runs across Meta, Reddit, and Google, targeting buyers at the person level via tools like Vector. All LinkedIn traffic routes to hyper-personalised microsites, not landing pages with a swapped-in logo, but pages where every line is written specifically around what that account cares about.

They do direct mail in two tiers. Volume-based plays (gift cards for demos) to reach users and develop account hypotheses. And bespoke, researched gifting via Wildcard to break through to VP and C-suite level at the enterprise accounts that matter most.

They chase outages, not generic intent. This is Kevin's "signal alpha", the competitive moat. His team scrapes Reddit and Downdetector for service degradations at prospect companies. An outage at a prospect's company creates genuine urgency around the kind of infrastructure tooling Datadog sells. He advised signals that are bespoke to your business are wide open.

The results:

• 25+ meetings per week with Fortune 500 companies
• SDR meeting quotas broken mid-year — too many meetings being booked
• 25% improvement in SDR activity efficiency within ABM accounts vs. standard motion

What they agreed on in Q&A

A few things that came up in the live Q&A are worth pulling out:

On build vs. buy: Kevin was candid, building yourself means duct tape. He hit 10 million cells in Google Sheets. The scalability and automation simply aren't there without purpose-built infrastructure. Mark's framing: the difference between cutting a garden path and building the interstate. Personal AI tools you vibe-code for yourself? Go for it. But the system that drives the numbers on your board slides needs to be hardened by experts. "Does anyone remember when every company tried to build their own CRM? Zero of those projects are still running."

On AI messaging: Both pushed back on over-engineering it. Mark: stop optimising for your own taste. You don't know what buyers will respond to until you send it, and most teams never send because they're too busy editing. Kevin: keep it under 300 characters, scannability wins, and if you're building with frontier models, test edge cases obsessively and remember garbage context in, garbage output.

On what flopped before they found what works: Mark: the disease of more. More calls, more emails, more accounts. It's the easiest default and almost never the most effective path. Kevin: broad-based advertising is dead. Running LinkedIn display campaigns to large audiences with no targeting is too competitive and too noisy. "If your outbound is hyper-targeted, your advertising has to be too."

The takeaway

Both Mark and Kevin arrived at the same place from different directions. They've figured out how to connect multiple signals into a single intelligence layer and use it to make their people more precise, not just more active.

That's the brain they both built. And it's working.

‍

Brains of GTM is UserGems' ongoing series featuring revenue leaders who are running signal-based, AI-powered GTM plays at scale.

Watch the full recording of this session.

‍