LLM commercialization has reached a fork in the road.

On July 15, Anthropic, together with Blackstone, Goldman Sachs, Hellman & Friedman and others, officially launched Ode, an enterprise AI implementation firm. The model is simple: deploy Anthropic engineers inside enterprises to help them actually make AI work. Not selling API call volume, not selling licenses — but embedding engineers, delivering projects, charging by results.

That sounds just like traditional IT consulting, right? Isn't that what McKinsey, Accenture, and IBM do? Yes — but the difference is: this time, the LLM vendor itself is getting into the delivery business.

Why Are LLM Companies Doing Delivery Themselves?

Because the biggest bottleneck in AI adoption has never been model capability — it's implementation capability.

Over the past three years, enterprises have spent a ton on AI — cloud compute, APIs, SaaS tools. But looking back, the percentage of projects that actually generate business value is depressingly low. A Gartner report last year said 85% of enterprise AI projects fail to meet expected ROI. Why? Not because the models aren't good enough, but because nobody knows how to integrate models with business processes.

Say you want to use AI for customer service. How do you connect it to existing systems? How do you ensure answers are accurate and not hallucinated? How do you hand off to human agents? How do you do quality monitoring? These questions aren't answered in LLM vendors' technical docs, and traditional IT consulting firms don't understand the boundaries of what LLMs can do. There's a huge gap in the middle.

Anthropic built Ode to fill that gap. Ode's engineers understand both Claude's capabilities and enterprise business processes — they're like "translators," converting business needs into tasks the model can understand, and translating model capabilities into implementable enterprise solutions.

The deeper reason: pure API sales have too low a growth ceiling. There are only so many teams in the world that can build with LLMs themselves. API revenue can only go so high. But how big is the enterprise AI implementation services market? The global IT services market alone is over $1 trillion. If AI implementation captures just 10% of that, it's a $100 billion market — way bigger than selling model APIs.

But the Challenge Is Obvious: Where Do the People Come From?

The model sounds great, but there's a practical problem: where do all the engineers come from?

Enterprise AI implementation is a labor-intensive business — you can only serve as many clients as you have engineers to deploy. Anthropic's own engineering team is only a few thousand people total; they can't all be pulled into delivery work. Blackstone and Goldman Sachs can easily invest money, but investing people is harder.

So whether Ode's model works depends on one thing: can it use AI itself to amplify implementation efficiency? If every Ode engineer, aided by agents, can serve 5–10 clients simultaneously, the business can scale. If it's still the traditional consulting "man-month" model, its ceiling is just another Accenture, not some new species of the AI era.

Either way, Anthropic has started something. LLM companies used to say "we're just a platform, delivery is for partners." Now Anthropic is saying "partners are too slow, we'll do it ourselves." It's a strong signal — the center of gravity in AI commercialization is sinking from the model layer down to the implementation layer. Whoever can help enterprises actually make AI work will be the one to make the next wave of big money.

See you tomorrow.