Most of GTM is glue work

The job you were hired for is a thin band of judgment. Everything around it is grind. That grind is the part to give away.

I was about to text 50 people.

Boot Camp cohort, five days to a hard close, warm leads scattered across four pools. An Impact Hub event from April. A partner community I was about to ask a friend to forward to. A German session from the Wednesday before. A private list of five or ten people I actually know. The instinct was a long list of names and a long evening.

Then I drew a grid. Four rows for the channels, two columns: broadcast and individual. Each cell asked one question. What fills this, here.

The grid answered in about a minute. The private pool has no broadcast option, only individual outreach. The partner community has a broadcast layer, the forward email, but the individual layer belongs to my friend, not me. Both of those gaps would have surfaced halfway through the evening, mid-execution, when they are expensive. On paper they surfaced before I sent a single message.

That is the part nobody tells you about go-to-market work. The thinking took a minute. The grind would have taken the night.

The thirty percent problem

Salesforce runs a recurring study on how salespeople actually spend their week. The number that keeps showing up: reps spend less than a third of their time selling. The rest goes to research, CRM updates, internal meetings, and chasing bad data.

Sit with the shape of that. The job is selling. Selling is the minority of the week.

It is not only sales. Data teams have a blunt name for the other half of their own job. They call it data-janitor work, the cleaning and shaping that has to happen before any actual analysis. Anaconda’s State of Data Science survey has put that at close to forty percent of their time, year after year.

The pattern holds across every operator role I have watched up close. Your real job is a thin band of judgment. Who matters. What to say to them. When to walk away. Around that thin band sits a thick layer of mechanical work: finding the data, shaping it, pushing it somewhere, cleaning up what broke. The layer is not the job. The layer is what stands between you and the job.

For two years the answer to that layer was to hire someone, or to buy another tool, or to stay late. There is a third answer now, and it is the one I want to walk through. When your work lives as files instead of trapped inside apps, an AI that reads files can take the layer off your hands. Not the judgment. The layer around it.

It shows up in three places: before your tools, in the middle, and after.

Before: the targeting

The grid is a “before” moment. Targeting, scoping, deciding where to point your effort before you spend any.

Most operators do this in their head, or in a doc that goes stale. The cost is invisible until you hit the gap mid-campaign. I did not draw that grid by hand, I built it with Claude Code against my own notes on who was in which pool. Two minutes. The output was not a prettier to-do list. It was a structural map that caught two holes I would otherwise have found the hard way, at nine in the evening, with the close clock running.

The work that pays off most in GTM is rarely the work that feels productive. Drawing the map feels slow. Sending the messages feels like progress. The map is what makes the messages land though.

The middle: getting faster AND smarter

A prospect once asked me what I had actually built. Fair question, and a surprisingly hard one to answer well.

I dispatched five agents in one batch, each pointed at a different region of my own files: skills, tools, frameworks, playbooks, products. They worked at the same time and reported back. I read the result the way the prospect would, noticed I had undercounted the client work, and sent a sixth agent at just that folder. The list roughly doubled.

Five minutes for an inventory that, done by hand, is a week of digging. The point is not the speed. The point is that I went into that conversation knowing my own position better than I had the day before. The middle layer, the part where a tool like Clay enriches your records or your CRM holds your pipeline, is exactly where this compounds. The machine does the fetching and the shaping. You arrive at the decision already smarter, not just earlier.

That is the half the demos never show. Everyone wants to see the thing that goes faster. The quieter win is the thing that makes you sharper before you open your mouth.

After: the glue, and the silent leaks

The after layer is the least glamorous and the most dangerous, because it fails quietly.

I went looking through my inbox for inbound leads. What I found instead: eighteen of thirty-two emails in a sixty-day window were my own out-of-office auto-replies. The auto-reply said I was out of reach until a date months away. It had been firing on real inbound. A client follow-up. People reaching in. A tripwire I set once and forgot, rejecting prospects on my behalf for weeks, and I could not see it because nothing tells you when your own glue breaks.

That is the after layer. Push to the CRM, trigger the follow-up, route the reply, log the result. None of it is hard. All of it is mechanical. And every mechanical step you cannot see is a place your funnel leaks without a sound. An AI that can read your actual setup, your filters, your replies, your routing, finds those leaks in the time it takes to ask. By hand, you find them when the quarter is already gone.

What stays yours

The move is the opposite of replacing your stack.

Claude Code does not cancel your tools. It lives on both sides of them. It does the un-sexy prep before your data hits Clay and the grunt-glue after it leaves, the scraping and shaping and pushing and cleanup that no one wants and everyone pays for in hours. Your tools keep doing what they are good at. The grind around them is what comes off your plate.

And the thin band in the middle, the judgment, stays exactly where it was. Deciding who matters. Choosing what to say. Knowing when the answer is no. The machine takes the layer that was never the job. What remains is the part that was always the point.

The question was never how to do the grind faster. It is what you would do with the week back.

If you are in Zurich on Wednesday, that is the whole conversation I am having at Clay Club with Lucas Maliczak. Claude Code on both sides of your stack, the work it quietly absorbs, and the judgment it hands back to you. Come build with us.