Welcome back to GTM Exclusives. Every two weeks we take a recent news story and give you the GTM angle the coverage misses.
This time: the people who build the AI models spent this year buying something else entirely, and that tells you where they think the value really sits.
Our data says they aren't the only ones. By the end of this, you'll have something specific to act on, whether you sell AI, sell into the industries adopting it fastest, or you're building your own stack.
What the announcements tell us
Four of the biggest names in AI spent this year committing capital to implementation.
Anthropic, with Blackstone, put $1.5 billion into a venture built entirely around getting AI working inside businesses. Microsoft stood up Frontier, $2.5 billion and 6,000 people who sit directly inside customer organizations. AWS committed $1 billion to a team that embeds engineers with clients. OpenAI is buying implementation firms outright to grow its own bench.
That's more than $5 billion from the three that put a number on it, and all of it points in the same direction: the hard part is no longer building the actual AI model.
Model prices have fallen, and results haven't followed, because what slows companies down sits in their data, their context, and their access to people who can connect the two. The named AI giants have now priced that work at billions.
What the announcements don't tell us
The announcements tell you that four companies made a decision. What they don't tell you is who else is making it, what they're building on, or where it's moving fastest.
So we went looking in our own data. If value really is moving toward implementation, it should show up in two places long before it reaches anyone's revenue: who companies are hiring, and what technology they're putting to work.
Three findings stand out.
1. There's a significant hiring surge for engineers with Gen AI skills
We track hiring surges, the moments when a company suddenly opens far more roles in a department than it usually would, and we can see the exact skills and tools those roles ask for.
The share of those surges demanding hands-on generative AI skills has climbed every single quarter, from 5.9% at the end of 2025 to 7.6% by the middle of 2026. Roughly 1 in 13 surging companies today, against 1 in 17 six months ago.
Two-thirds of every surge we tracked sat in engineering and technical teams, which tells you what kind of work is being staffed.
And there's one interesting detail that matters more than the headline number itself. Overall surge volume rose and fell across the period, as hiring always does. The AI-building share of it only went one way.
What that means for you: a generative AI hiring surge is a buying window that opens before procurement does. Budget has been signed off, a build has started, and the tooling decisions are still live. Most teams find these accounts months later, by which point the decision has been made, and somebody else is already in the room.
They're hiring to build. The next question is what they're building on.
2. What they're building on changed inside six months
The same roles tell us which tools companies expect to build with, because they name them.
Mentions of Claude in surge hiring went from 33 in Q4 2025 to 125 in Q2 2026. Close to four times as many in two quarters, and enough to make it the most requested AI tool in these roles.
Set aside which vendor is ahead for a moment. The really useful fact is that a leaderboard can move this far in six months. Whatever ranking you had in your head at the start of the year is already describing a different market.
Some context for why it moves so fast: the AI software market your buyers are choosing from runs to 1,492 vendors in our own technology index, and 92% of them have exactly one product in it.
What that means for you: the tool named in a job description tells you which stack a company is committing to, and it tells you before the announcement, the case study or the press release. It also has a short shelf life. If your integration roadmap was set twelve months ago, it was built for a market that's moved underneath it.
3. Adoption is up 38%, and the fastest growth is nowhere near Silicon Valley
The final signal is adoption: how many companies have actually put AI and machine learning technology into their stack.
We track 1,835 named AI and ML products, from the major platform vendors down to single-product specialists, across every business function from marketing to security. Adoption is up 38% in six months, measured from the end of 2025 to the middle of 2026.
The heaviest adopters are the firms doing the implementing. Custom software and IT services companies are adopting faster and in greater numbers than anyone else, up 52%. These are the firms clients hire to run AI projects, and they're buying the tooling before their clients do.
Then the part that should influence how you think about your target list. The fastest-growing adopters are old-economy industries:
Construction is up 72%
Financial software, 87%
Insurance, 76%
Law firms, 72%
What that means for you: almost everyone selling AI right now is pointed at technology companies. That's the most contested part of the market, and on this data it's some way behind the fastest-growing part of it.
What the hiring and adoption data says about where this lands
Put the three findings next to the announcements, and the story somewhat changes shape.
The giant announcements are a lagging indicator. Those decisions were taken months before anyone published anything, and by the time they did, hundreds of companies had reached the same conclusion. The giants actually announced it — you can only see everyone else by reading their job ads.
And there's a second reading of the same numbers. The giants are buying implementation capability. Companies are hiring for implementation capability. Everyone with capital is bidding for the same scarce thing, and it's people who can connect a model to a business.
The two signals also sit at different points on the same timeline. A hiring surge tells you a build is being staffed, which puts the company months away from having anything running. Adoption tells you the technology is already in the building. Treat them the same, and you'll pitch a company months from being ready, or reach one that already decided.
Both of them move ahead of revenue. A company hiring to build hasn't bought yet. A company that's just put AI technology into its stack hasn't shown a return yet.
The concentration is something that really surprised me. If this were simply a story about AI maturity, technology companies would lead it, but they don't. The fastest movement is in construction, insurance, financial software and law, industries that don't fit the usual picture of AI adoption.
Which brings us to the part that matters most, depending on where you sit.
What this actually means
If you sell AI, data or implementation services
Your target list could be pointed at the wrong industries. The list that looked obvious 18 months ago is now the one almost everyone selling AI is working. Construction is adopting at 72% growth, for example.
A job ad reaches you earlier than intent data. A company advertising for hands-on generative AI skills has already signed off the budget and started building. Intent data picks them up later, once they start comparing vendors, by which point the requirements are usually set.
Read the tool names. The stack a company names in its job descriptions is the stack it's committing to. That list moved inside six months, so treat what you know about it as perishable. If your integration story was built around whichever tool led last year, check it still matches what buyers are asking about.
If you sell into construction, insurance, legal or financial services
Your market is changing faster than the coverage suggests. These four are the fastest-growing adopters in our dataset, and that growth is already behind them rather than forecast ahead of them. If your account plans assume construction or insurance moves slowly on technology, those plans were written for a version of the market that has quite literally gone. The buyers you've been pacing yourself against have started moving, and they did it within the last six months.
The list is easier to win, and the play is harder to run. Fewer competitors are calling these accounts, which helps. The rest of it is that you are selling into different buying committees, longer cycles, and rooms with less AI literacy in them.
The capability pitch that lands in tech falls majorly flat here, because the question in the room is whether this works in a business like theirs. That means a named reference in their own industry does more than a feature list, and a first scope small enough to prove out in a quarter beats a transformation program nobody wants to sign for.
If you're building your own GTM AI stack
The giants made your argument for you. The organizations with the best possible access to frontier models spent billions on the layer between the model and the business. If the people who build these systems think that's where the difficulty sits, it's worth taking seriously in your own planning.
Everyone can buy the same model. Your competitors have access to exactly what you have. What they don't have is your first-party data: what your customers actually did, said and bought. The open question is whether it's current enough, and connected enough, for an agent to act on it without someone checking the output first.
One question to take into planning. What would have to change for an agent to answer a live customer question using your data, with nobody verifying it beforehand? Whatever comes back on that list is the implementation work the giants just priced at billions.
The final word
An announcement tells you what one company has decided. It rarely tells you how many others reached the same conclusion without saying so, and that's usually where the commercial opportunity sits.
Everything above came from tracking hiring surges and the adoption of 1,835 named AI and machine learning technologies between the end of 2025 and the middle of 2026, then reading the two together.
Four companies announced their investment. The rest of the market is doing the same, but in job postings that can be missed as a signal.
I'll see you again in two weeks, when we'll decode another GTM news story.
— Dennis
