The short version: In 2026 both frontier labs placed the same bet in two different corporate shells — that the money in enterprise AI is in implementation, not model quality. OpenAI launched its majority-owned Deployment Company on May 11 with more than $4B from 19 investors (TPG, Advent, Bain Capital, and Brookfield co-leading) and staffed it by acquiring the consultancy Tomoro and its roughly 150 forward-deployed engineers. Anthropic's version, Ode with Anthropic, launched July 15 as a $1.5B venture with Blackstone and Hellman & Friedman, built on the acquisition of Fractional AI, fielding about 100 embedded engineers aimed at CEO-level transformation — Claude-first but not locked to Anthropic. If you sell "we'll integrate AI into your business," your buyer can now hire the model-maker to do exactly that. The moat moved from tokens to the last mile.
The two ventures, side by side#
The structures are more different than the headlines suggest.
OpenAI's Deployment Company is a majority-owned, OpenAI-controlled subsidiary — reported at a valuation somewhere between $10B and $14B — capitalized with over $4B from a 19-investor consortium. It went from zero to a staffed services firm by buying Tomoro, a London-based applied-AI consultancy founded in 2023 in alliance with OpenAI, which brought roughly 150 Forward Deployed Engineers and deployment specialists on day one. McKinsey, Bain & Company, and Capgemini are named on the integration side.
Ode with Anthropic is a joint venture, not a subsidiary. Anthropic is one partner alongside Blackstone and Hellman & Friedman, with Goldman Sachs a founding investor and Apollo, GIC, Sequoia, and General Atlantic also in. It was built on the acquisition of Fractional AI — a firm founded in 2024 by Chris Taylor, Eddie Siegel, and Travis May (all ex-LiveRamp), now renamed Ode and led by Taylor as CEO and Siegel as CTO. Notably, that acquisition reportedly pulled Fractional AI out of OpenAI's orbit. Ode's roughly 100 engineers were described by its backers as "special forces" rather than a large army, with more than half former founders.
One detail worth correcting, because it's easy to get backwards: only Ode is built on Fractional AI. OpenAI's venture runs on Tomoro. Both are staffed by acquisition; different acquisitions.
The bet underneath both deals#
Strip away the cap tables and both moves say the same thing: the frontier model is now table stakes. When GPT-class and Claude-class systems are close enough that most enterprises can't tell the difference on their actual workload, the model stops being the product. What's scarce is the engineering that turns a model into a working system inside a real company — the integrations, the data plumbing, the evals, the change management, the accountability for an outcome.
That's the forward-deployed engineer thesis, and it's a tell. Labs don't stand up multi-billion-dollar services arms and buy consultancies unless they've concluded that value is leaking past the API into the implementation layer — and they'd rather capture it than watch a fragmented services market do it. Ode's own framing is that model selection matters but isn't where most of the work happens; the work is engineering the full system around the model. OpenAI's is nearly identical: embed engineers who turn AI gains into "durable systems."
This is the same current we tracked in July's ~$1.8B agent-funding wave, where the biggest checks skipped the model labs and went to companies governing and verticalizing agents. And it's item four in this week's Founder's Wire: nobody shipped a new flagship, and the leverage moved to everything around the model.
What actually changed for founders#
Two things, and they pull in opposite directions.
The bad news: "We're AI-powered" is now worth roughly nothing as a pitch, and "we'll integrate AI for you" is worth less than it was in the spring. Your enterprise buyer can hire OpenAI's or Anthropic's own engineers to build the integration. Generic implementation — prompt plumbing, a RAG pipeline, a chatbot over their docs — is exactly the commodity these ventures are built to deliver at scale, with the model-maker's brand behind it.
The good news: both labs are aiming at the top of the market — nine-figure transformation programs, CEO-sponsored, private-equity-backed rollouts. Ode explicitly prioritizes accounts with executive commitment. That's not where a small builder competes, and it's not where most of the market lives. The labs' FDE teams are expensive, general-purpose, and allocated to the biggest logos. Everything below that line, and everything too specific for a generalist to touch, is still open.
What to do#
Pick the last mile and own it. Concretely, this week:
- Audit your pitch for the word "AI." If your differentiation is "we use GPT/Claude," rewrite it around an outcome your customer can measure. The model is now assumed, not sold.
- Find the part a lab's FDE won't specialize in. That's usually one of: a narrow vertical with real domain rules, proprietary data or workflow you can access and they can't, a regulated process where you'll carry the liability, or ongoing accountability for a result over months. Anchor there.
- Stay model-agnostic on purpose. Ode itself is Claude-first but not locked, and switches models when the system design demands it. If a $1.5B venture won't chain itself to one model, neither should you — it keeps your leverage and your costs honest.
- Sell integration as an outcome, not a technology. The FDE model works because it owns results, not deliverables. Price and position the same way, at a scale the labs' teams will never bother to serve.
The labs just told you where they think the value is. Read it as a map: they've claimed the frontier model and the biggest transformation deals. The domain, the data, and the last mile are still yours to defend.
