Anthropic and OpenAI both announced today that they're standing up implementation arms. Anthropic launched a $1.5 billion joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs, designed to embed Anthropic's applied AI engineers inside client organizations. OpenAI finalized a $10 billion vehicle anchored by TPG, called The Deployment Company, that will put teams of OpenAI engineers directly inside enterprise clients. Two of the most valuable AI companies in the world just told the market, in the most public way they could, that the technology by itself is not enough. Someone has to do the work of putting it into a business.
What Was Said
The framing from both companies was similar. The models are capable. The frontier is moving fast. But the gap between what the technology can do in a demo and what it actually delivers inside a real organization is wide, and closing it requires people on the ground. Forward-deployed engineers, in OpenAI's language. Implementation specialists, in Anthropic's. The job titles vary. The work is the same.
Aaron Levie described that work in detail in a post that took off on X yesterday. He walked through what it actually takes to put agents into a real organization. You have to get them talking to your data securely across legacy systems that were built decades ago. You have to implement the right access controls and entitlements, and you have to monitor and log what the agents do. You have to document the processes that have lived as tribal knowledge for years. You have to redesign the workflow so people and agents are working together in a way that captures the gain, rather than just replicating the old process with an agent bolted on. And you have to build evals for the new end-state, then keep up with a set of best practices that's changing every few weeks.
By the time Levie's post had been up for a few hours, it was the centerpiece of a trending topic on X with a Grok summary headline of "AI Agents Demand Far More Enterprise Work Than Expected." That summary cited a March 2026 survey showing 78% of firms are piloting agents and fewer than 15% have reached production. The bottleneck is not the model. It's everything around the model.
Why This Matters
When the model providers themselves stand up consulting arms, that signals something specific about where the value is.
For most of the last two years, the public conversation has been organized around model capability. Which model is best, which one passes which benchmark, which one can write the cleanest code. That conversation is real, and it matters. But it has obscured a more practical question: what does it take to turn any of this capability into business outcomes?
The answer is now coming back loudly, from multiple directions. It takes a translation layer. It takes someone who can sit between the tools and the operations of a real company, understand both, and do the unglamorous work of connecting them. The legacy data has to be made legible to the agent. The processes have to be written down. The access controls have to be respected. The workflows have to be redesigned. The people have to be trained. None of this is exotic. All of it is required.
Anthropic and OpenAI building out their own arms is the most credible confirmation yet that this layer is real, and that the work to do it is enormous.
Why We Agree
We've been operating in this space for a year. The premise of the firm has always been that the gap between AI capability and AI outcomes is closed by people who can sit inside an organization, not by people who can write better marketing copy about it. Embedded work. Department by department. Workflow by workflow. The kind of thing that doesn't show up in a benchmark and doesn't fit on a slide.
What's changed today is that the largest model companies in the world have publicly endorsed that premise with their balance sheets. That's good news for any operator who's been quietly investing in this work, and it's especially good news for the companies who need that work done.
How the Market Is Shaping Up
It's worth being clear about what these announcements probably mean for the structure of the market.
The model companies will not own all of this work. They won't try to. Their economics will push them toward the largest enterprises in the world, the ones with budgets that justify a multi-year, multi-team engagement and a long bench of forward-deployed engineers. Both announcements today made that explicit. Anthropic's joint venture is initially aimed at the portfolio companies of Blackstone, Hellman & Friedman, and the other PE participants. OpenAI's is anchored by TPG. The labs are setting up to serve the highest-value end of the market, where a single deployment pattern can be replicated across dozens of similar businesses. They'll function the way the global consultancies have always functioned in adjacent categories. They'll set the high-end pattern, win the marquee logos, and define what "good" looks like at scale. That's a useful thing to have in any market. It legitimizes the category and pulls expectations up.
Below that tier, the market looks different. There are tens of thousands of companies, somewhere between fifty employees on the low end and several thousand on the high end, that need exactly the same translation work and will never be a target account for OpenAI's or Anthropic's implementation arm. The economics don't line up for either side. Those companies don't have the budget for an MBB-scale engagement, and most of them don't have the internal AI engineering bench to do this work themselves. They need firms that can come in, embed, do the real work, transfer the capability back to the team, and leave them better equipped than they were before.
That layer is where most of the actual translation work in the economy is going to happen over the next few years. It's a different kind of firm than the ones building model APIs, and a different kind of firm than the ones writing thousand-page transformation decks. It's smaller, more embedded, more operational, and more focused on what an actual department needs on a Tuesday morning.
What Comes Next
The companies that get the most out of AI in the next two years won't be the ones with the most expensive tools or the boldest internal slogans. They'll be the ones that did the careful, embedded, unglamorous work of putting AI into their actual business. That's true whether the work is done by OpenAI's new arm, an MBB firm, a smaller consultancy, or a team built internally.
What the announcements from Anthropic and OpenAI confirm is that this work is real, that it's a category, and that it's worth investing in seriously. Plenty of people in the market already knew that. It's a healthier market when the largest players are saying it out loud.
If any of this is on your radar and resonates, we'd love to talk. It's the work we do at Gadoci Consulting.