On Tuesday, Anthropic and Blackstone put a name on something they'd been building quietly since May: Ode, a $1.5 billion company whose entire job is helping businesses actually use AI. Not build models. Not sell subscriptions. Just wire the technology into how a company already works.
They're not alone. OpenAI spun up its own version, called The Deployment Company. Microsoft committed $2.5 billion to the same idea earlier this month. Deloitte and Accenture both launched "forward-deployed engineering" practices this year. Goldman Sachs and Hellman & Friedman are backers on the Ode deal. That's a lot of serious money agreeing on one thing at the same time.
Pay attention to what this actually says. The companies that make the AI models — the ones with every incentive to tell you the next model will solve everything — are betting billions that the hard part isn't the model at all. It's getting the model to do real work inside a real business.
The quiet admission in the loud announcement
Ode's chief technologist put it plainly: model selection is "not where the majority of calories are spent." He compared choosing a model to choosing a programming language — one ingredient in a system that has to be engineered around your actual operations. The value isn't in the tool. It's in the wiring.
This matches what the small-business data has been saying all year. Roughly 61% of small businesses now use AI, and another 29% are experimenting — but close to one in three of the businesses using it report being stuck in the testing phase, never quite getting from "we tried it" to "it runs part of the business." The tools got better every quarter. The stuck number barely moved. That gap is exactly what these billion-dollar companies exist to close.
When the model makers spend billions on implementation, they're telling you the model was never your bottleneck. The wiring is.
You can't hire Ode. You don't need to.
Ode's ideal customer is a company where the CEO has made AI the top priority for the next two years — think Blackstone portfolio companies, not the local med spa. A hundred elite engineers aren't coming to your office, and their day rates wouldn't make sense for you anyway.
But the playbook they run is not secret, and it scales down further than you'd think. Strip away the enterprise language and it's three steps. Pick one process, not ten. Ode works on the single most important workflow a company has, not a grab bag of experiments. For you that might be lead follow-up, review replies, or your weekly content. One. Engineer around your operations, not the tool. The work is writing down how the process actually happens today — who does what, where the info lives, what "done" looks like — and then fitting AI into that. Most stuck businesses skipped this step and started with the tool. Measure the result. Ode runs constant evaluations to prove business impact. Your version is simpler: hours saved per week, or replies sent that used to slip. If you can't point to a number after a month, unwind it and pick a different process.
That's the whole trade. The businesses stuck in experimentation mode aren't stuck because their AI is too weak. They're stuck because nobody sat down and did the unglamorous wiring work — the part the biggest companies in the world just decided was worth billions.
The window this opens
Here's the part worth sitting with. Big companies are about to get this done for them, at enterprise prices, over multi-year engagements. Small businesses can do the scaled-down version themselves, this quarter, for close to nothing — the models are cheap now, and the process work takes an afternoon per workflow. For once, being small is the advantage: you don't have legacy data infrastructure to untangle. You have one messy process and the person who understands it is reading this.
Stop waiting for a smarter model. The people who build them just told you it won't be the thing that changes your business. The wiring will.