The Underdog Just Won the Enterprise. Here Is Why Your Vendor Choice Was Never Safe.

Anthropic has passed OpenAI in US business adoption for the first time and launched a partner program to lock in the companies building on Claude. The headline is the leaderboard. The real story is that the company you standardized on three years ago is no longer obviously the right bet, and most enterprises built no way to change their mind.

For three years, the AI vendor question felt settled. A company piloted one model in 2023, liked it well enough, wired it into a few workflows, and moved on to the next problem. The decision got made once and then quietly hardened into infrastructure. Nobody revisited it, because revisiting it sounded like make-work.

This week that assumption cracked. Anthropic surpassed OpenAI in US business adoption for the first time, and it did not stop at the milestone. The company launched the Claude Partner Hub and a Services Track, formalizing a $100 million partner program aimed at the firms that put Claude into production. The move follows a $65 billion Series H that valued Anthropic near $965 billion post-money, a number that briefly made it the most valuable AI company on the board, ahead of OpenAI's last private round.

I advise large companies on how to actually implement this technology, and I want to be clear about what is and is not interesting here. The leaderboard is not interesting. Anthropic and OpenAI will trade the top spot more than once before this is over, and so will Google, and possibly a name none of us are saying yet. What is interesting is the structural lesson hiding inside a vendor's good quarter.

The decision you made once was never meant to last

Most enterprise AI decisions were made under a false assumption: that the frontier would settle down. Executives picked a vendor the way they pick a payroll system or a CRM, expecting it to be a ten-year choice. They negotiated the contract, integrated the API, trained their teams on one model's quirks, and treated the work as finished.

But AI is not a payroll system. The frontier moves on a roughly ninety-day cycle. In the time it has taken some companies to roll out a single model across departments, the field has produced new flagship releases from every major lab, including Claude Opus 4.8, OpenAI's GPT-5.5 line, and Google's Gemini 3.5 Flash, each resetting the benchmark the others chase. The thing you standardized on is rarely the best thing available by the time your rollout finishes.

This creates a trap that has nothing to do with which lab is winning. The trap is that you built no mechanism to change your mind. If switching from one model to another would take your engineering team a year of rework, you did not buy a tool. You bought a cage with a logo on it.

Why the partner program matters more than the funding

Look closely at what Anthropic actually did this week. The funding round is the headline, but the partner program is the strategy. A $100 million investment in partners who implement Claude in production is not generosity. It is an effort to make Claude sticky, to wire it so deeply into the consulting firms, system integrators, and internal teams that move enterprise software that ripping it out becomes unthinkable.

This is the same playbook that made enterprise software a fortress for decades. You do not win the enterprise by being the best product in a given month. You win it by becoming the default that nobody has the budget or the appetite to replace. Every major AI lab now understands this, which is why the competition is moving from raw capability to lock-in.

For the leaders watching this, the takeaway runs opposite to the vendor's intent. The more aggressively the labs compete to lock you in, the more valuable it becomes to stay portable. The companies that will get the most out of AI over the next three years are not the ones that pick the right model today. They are the ones that build an abstraction layer between their workflows and any single model, so that when the frontier moves, they can move with it in a weekend instead of a fiscal year.

The reshuffle is a feature of the market, not a glitch

There is a tendency among executives to read every vendor change as instability they need to wait out. That reading is backwards. The reshuffle at the top is not a temporary mess that will resolve into a clear winner you can finally commit to. It is the permanent condition of this market. Analysts now describe a reshuffling of model allegiances as one of the defining forces in enterprise AI for 2026, alongside a cost crisis and a security exposure that grows with every autonomous agent deployed.

If the allegiance reshuffle is permanent, then waiting for it to stop is a strategy of waiting forever. The competent response is not to predict the winner. It is to design your operations so the winner does not matter much to you, because you can adopt whoever is best this quarter without tearing anything down.

This is where the beekeeper view helps. The job of a leader in an AI-driven company is not to manage a particular set of bees. It is to direct whichever bees are strongest toward the outcome you want. If you fall in love with one swarm and rebuild your whole hive around its specific behavior, you have made yourself fragile to the one thing you can count on, which is that a stronger swarm is coming.

What to actually do this quarter

Start with the contract. Pull up your primary AI vendor agreement and find the switching cost, not the dollar cost, the operational one. How long would it take, in engineer-weeks, to move your core workflows to a different model? If the honest answer is more than a month, you have a portability problem that is more urgent than any capability gap.

Then look at your architecture. Are your applications calling a specific model directly, or are they calling an internal interface that could route to any model? The first design makes the vendor's lock-in strategy work on you. The second one makes their competition work for you.

Finally, set a review cadence. The single most common mistake I see is treating the model decision as permanent when the market treats it as quarterly. Put a recurring date on the calendar to re-evaluate which model serves each major workflow best. You will not switch every quarter. But you will know your options, and knowing your options is the entire difference between negotiating from strength and renewing from inertia. The teams that run this cadence rarely make a dramatic change, yet they consistently get better terms and adopt improvements months ahead of peers, simply because they never let the decision go stale. Awareness is cheap. Ignorance compounds.

The cost of waiting is invisible until it is enormous

One reason this trap catches so many capable leaders is that the cost of vendor lock-in does not show up on any report until it is too large to fix quickly. There is no line item called falling behind. The dependence accumulates quietly, one integration at a time, until the day a clearly superior option arrives and you realize you cannot use it without a project you should have started a year ago.

By then the competitor who stayed portable is already shipping with the better model, and you are still scoping the migration. The gap that opens in those months is not recoverable on the same timeline, because while you rebuild, they keep moving. This is how a market with a ninety-day frontier punishes commitment to any single supplier: not with a sudden failure, but with a slow widening of the distance between the companies that can adopt and the companies that can only watch. The bill for a decision made once in 2023 comes due in a quarter you did not choose, at a size nobody forecast, and it is paid in lost ground rather than dollars, which is exactly why it is so easy to ignore until it is severe.

Portability is a negotiating position, not just an engineering one

There is a commercial dimension to this that gets missed when the conversation stays technical. The companies locked into a single model are not just slow to adopt better technology. They are weak at the negotiating table.

Consider what happens at renewal. If your vendor knows it would take you a year to leave, the price conversation is over before it starts. They hold all the leverage, because your alternative to their terms is a year of disruption you cannot stomach. Now consider the company that built a routing layer and can move core workflows to a competitor in a week. That company walks into the renewal with a real alternative, and a real alternative is the only thing that has ever produced a real discount. Portability is not just insurance against falling behind on capability. It is the difference between being a price-taker and a price-setter in a market where the cost of switching is the whole negotiation.

This is why the reshuffle at the top is good news for buyers who prepared and bad news for buyers who did not. Analysts list cost as one of the defining pressures in enterprise AI this year, with finance teams struggling to keep pace as spending climbs. The companies feeling that cost crisis most acutely are often the ones that committed early and deeply to a single vendor and now have no way to put competitive pressure on their bill. The labs raising money at near-trillion-dollar valuations are counting on exactly that dependence. Every dollar Anthropic puts into its partner program is a dollar spent making your exit more expensive, and the return on that spend shows up in your next invoice.

The leaders who understand this treat model selection the way a sophisticated buyer treats any critical supplier. You qualify multiple sources, you keep them competing, and you never let any one of them believe it is the only option you have. That posture costs a little more in engineering up front and saves a great deal in both money and freedom over time.

Anthropic passing OpenAI is a good headline and a better warning. The warning is not that you bet on the wrong horse. It is that you built a stable that only fits one horse, in a race where the lead changes every ninety days. So here is the question worth sitting with this week: if a clearly better model launched tomorrow, how fast could your company actually use it, and who in your organization can answer that without guessing?

Sharon Gai is an AI transformation strategist, keynote speaker, and author of How to Do More with Less Using AI. She advises Fortune 500 companies on AI adoption and organizational redesign.