A $7.1bn valuation for a job you haven't hired

In September 2026, Clay raised $115 million in a Series D at a $7.1 billion valuation. Wellington led it, with Sequoia, a16z, Meritech and CapitalG in the round. They now report over 17,000 customers, including 80% of the Forbes AI 50.

I'm not writing about the money. The money is the least interesting part.

What the money is actually buying is a bet on a job title. The GTM engineer. Somebody who sits between RevOps, data and sales, who can wire a list of accounts to a set of signals to a sequence, and who thinks in systems rather than in activity. Three years ago that person didn't have a name. Clay say they coined it. Now the market has put billions behind them.

They have also put $1 million into a scholarship fund to train people into the job. That's the part worth watching. You don't fund a training programme for a role unless you need that role to exist in enough companies to buy what you're selling.

Here is the problem. Almost nobody I speak to in the mid-market has that person. Most of them don't have a clean CRM either.

Seventeen Thousand Customers

Let's be fair to Clay before I start kicking. 17,000 customers is not vapour. Anthropic, Google, OpenAI and Stripe are on the list. That is a real product solving a real problem, and the problem is genuine: finding the right accounts at the right moment used to take a researcher a week and now it takes a workflow an hour.

Does that mean it'll work in a 40-person consultancy in Leeds? Probably not out of the box. Does it mean founders are about to be sold it anyway? Yes, 100%.

Because the thing about a $7.1 billion valuation is that it buys a lot of sales and marketing, and a chunk of that is going to land in your inbox over the next six months.

Swapping A Person For A Platform

One of my clients hired a physical BDR. Spent £110,000 on him across the year. Got zero leads.

Not "underwhelming pipeline". Zero.

Now, that wasn't the BDR's fault, and I'd say the same about any tool. He was dropped into a business with no written ICP, no agreed definition of a qualified conversation, and a CRM where the same company existed three times under three spellings. He did what you'd expect anyone to do in that situation, which is work hard in roughly the wrong direction for twelve months.

Swap that £110,000 person for a £1,500 a month platform and you don't fix anything. You just make the wrong direction cheaper and faster. That's the whole lesson and it's not a complicated one.

Clean Pipes First

I'd like to say this is just my hobby horse, but the vendors have started saying it too, which is the bit I find genuinely interesting.

LXA's 2026 B2B State of Martech and Revenue Operations report, produced in partnership with LeanData and surveying 201 senior B2B leaders at companies with 2,500 or more employees, found 82% agreeing that clean data and reliable routing have to come before scaling AI, and only 26% with the enforcement mechanisms to act on it. 42% named poor alignment on lead qualification as a significant gap. 32% reported duplicate or mismatched lead-to-account records. Process and operations scored lowest on their maturity index, and improved least of the five pillars over three years of benchmarking.

Fair warning on that one. It's enterprise respondents, not SMBs. I haven't been able to pin down an exact publication date beyond the April 2026 field period, so take the timing as directional. And LeanData sponsored research that concludes you need exactly what LeanData sells, which is routing and lead-to-account matching. Worth knowing before you quote it at anyone.

Two things make me keep it anyway. The sample is 35% UK and 25% US, which is closer to home than most of the research that gets quoted at British founders. And if companies with 2,500 staff and a dedicated ops function can't keep their records straight, I'd gently suggest the 40-person firm with a part-time CRM admin isn't quietly winning that fight either.

HubSpot went further. In their Fall 2026 Spotlight they launched something called Context Home, which gives your business a completion score for how good the context is that your AI is working from. Read that again. A major CRM vendor has conceded, in product, that the blocker isn't the model. It's the state of your records. And they've built a scoreboard for it.

HubSpot also claim customers with strong context create 3.6x more MQLs and win 3.2x more deals. Those are HubSpot's numbers, about HubSpot customers, footnoted to a source they haven't made public. I wouldn't put them in a business case and I'd be wary of anyone who does.

What I'd Actually Do

There's no silver bullet here, otherwise we'd all be rich and famous. But there are three things that are cheap, boring and have to be true before any of this tooling earns its keep.

Write down who you sell to. Not a persona deck. A list of the attributes that made your last ten good clients good, and the attributes that made the bad ones bad. If you can't write it, no tool can infer it.

Agree what a qualified conversation is, in one sentence, with whoever owns the number. 42% of those enterprise respondents named this as a significant gap. You're probably not the exception.

Then clean the records. Deduplicate, fix the account hierarchy, decide which fields are mandatory and actually enforce it. It's the least glamorous week of work you'll do this quarter and it's the one that makes everything after it possible.

Do those three and a GTM engineer, or Clay, or Breeze, or whatever comes next, has something to work with. Skip them and you've bought a very expensive way to annoy your pipeline faster.

The ink on the $7.1 billion is dry. Yours isn't.

Common questions

What is a GTM engineer? A GTM engineer sits between revenue operations, data and sales. They take a list of target accounts, wire it to a set of buying signals, and turn that into outreach that actually fires, usually by building workflows across the CRM and whatever enrichment and automation tools sit around it. The distinguishing thing is that they think in systems rather than in activity: the output is a repeatable mechanism, not a number of emails sent. The title barely existed three years ago and most mid-market firms still don't have anyone doing the job.

Should you fix your CRM data before buying AI sales tools? Yes, and the vendors have started admitting it. LXA's 2026 survey of 201 senior B2B leaders at companies with 2,500 or more employees, produced with LeanData, found 82% agreeing that clean data and reliable routing have to come first, with only 26% having the enforcement mechanisms to act on it. HubSpot now ships a completion score for how good the context is that your AI is working from, which is a CRM vendor conceding in product that the blocker is the state of your records rather than the model. If organisations that size can't keep their records straight, a smaller firm with a part-time CRM admin almost certainly can't either.

What does clean CRM data actually mean? Three concrete things, none of them glamorous. Deduplicate, so the same company doesn't exist three times under three spellings. Fix the account hierarchy, so subsidiaries, divisions and parent companies resolve to the right record. Then decide which fields are mandatory and actually enforce it, because a field nobody has to fill in is a field nobody fills in. It's about a week of work and it's the thing that makes everything you buy afterwards worth the money.