OpenAI acquired Tomoro.ai in May 2026 to anchor its $4B enterprise deployment arm. The leading independent applied-AI consultancy is no longer independent.
Bloomberg, May 11 2026

Tomoro.ai built something genuinely valuable: an applied-AI consultancy that helped enterprises navigate a technology landscape most organizations did not have the internal expertise to evaluate on their own. Their work was respected. Their independence was the point. When OpenAI acquired them to anchor a $4 billion enterprise deployment arm, it was not a commentary on the quality of their advisory. It was a commentary on the structure of the market.

The structure is this: AI consulting is consolidating into the platforms it deploys. The advisory layer and the vendor layer are merging. That is a structural shift with consequences for every enterprise that has hired or is considering hiring an AI advisory partner, and it deserves more attention than it is currently receiving.

When McKinsey recommends SAP, they disclose the relationship. The enterprise knows to weigh the recommendation accordingly. When an AI consultancy is owned by OpenAI, the relationship is not a disclosure footnote. It is the business model. The advisory firm's growth targets and the platform's growth targets are the same targets. The question of whether those targets align with the enterprise client's optimal outcome is a question the enterprise can no longer assume has been answered in its favor.

What vendor lock-in actually means in AI

SaaS lock-in is familiar territory. Data portability is constrained. Switching costs are high. Contracts auto-renew. The exit is painful enough that most organizations stay. But the lock-in is at least visible. The enterprise knows it exists and can price it into the decision.

AI lock-in operates differently, and in some ways more deeply. It has three distinct forms, each of which compounds the others.

Form 1
Model Lock-In
Fine-tuning on a proprietary model creates a trained artifact that does not transfer to another vendor's infrastructure. The more an organization fine-tunes on one platform, the more expensive it becomes to switch. Unlike a database migration, a fine-tuned model represents months of training data curation, labeling work, and evaluation effort that cannot be ported. The organization has built on top of a foundation it does not own and cannot move.
Form 2
Infrastructure Lock-In
Data pipelines built on vendor-specific APIs, embedding formats, and retrieval architectures create deep technical dependencies. A RAG system built on one vendor's vector store and embedding model is not straightforwardly portable to another. The connectors, the schema, the latency assumptions, and the evaluation framework are all shaped by the original vendor's architecture. Rebuilding costs time and erases institutional knowledge accumulated during the original deployment.
Form 3
Recommendation Lock-In
This is the least visible form and, in the current market, the most consequential. When the advisory firm recommending your AI architecture has a financial relationship with one platform vendor, the recommendations carry a structural bias that the enterprise cannot easily detect or correct for. The recommendation to use a particular RAG framework, fine-tuning infrastructure, or model family may be technically sound. Or it may reflect the path that best serves the platform's adoption metrics. The enterprise has no reliable way to distinguish between the two.

When your AI consultant is owned by a platform, their incentive is not your optimal outcome. It is platform adoption. That is not an accusation of bad faith. It is a description of how incentive structures work. The platform that spent $4 billion to acquire an enterprise consulting arm did not do so to give enterprises neutral advice about when not to use its products.

The consolidation trend

The Tomoro acquisition is not an isolated event. It is a data point in a pattern that has precedent.

In the 1990s, the Big Four accounting firms began acquiring strategy boutiques. The pitch was integrated advisory: financial, operational, and strategic advice from a single firm. For a period, it worked. Then the independence scandals arrived, and regulators concluded that auditing a company and simultaneously consulting for it created conflicts of interest that could not be managed by internal firewalls. The Sarbanes-Oxley separation requirements followed. The consulting arms were spun off.

The AI consulting consolidation follows the same structural logic, with two meaningful differences. First, it moves faster. The AI market does not have the luxury of a decade-long scandal and a regulatory response to establish the independence norms that eventually emerged in accounting. Second, the lock-in is technical, not merely relational. A conflicted audit opinion can at least in principle be challenged. A conflicted infrastructure recommendation produces systems that are expensive to reverse even after the conflict is identified.

McKinsey, BCG, and Accenture all have AI practices. All have AI vendor relationships, announced partnerships, and preferred technology ecosystems. The enterprise engaging any of them for AI advisory should ask: is this recommendation the product of an analysis that began with my problem, or an analysis that began with a vendor relationship? That question was always appropriate to ask. It is now more urgent than it was twelve months ago.

The boutique applied-AI firms that emerged over the last several years, Tomoro among them, were genuinely independent. They were small enough that platform partnerships were not structurally central to their business model. They made technology recommendations based on what the technology could do for the client. That independence is now being absorbed into platforms at a rate that will leave the market with very few firms that can credibly claim neutrality.

What vendor-neutral means in practice

Vendor-neutral does not mean vendor-agnostic. It is not a refusal to work with OpenAI, Anthropic, Microsoft, or Google. Every serious AI engagement involves at least one of them. The question is not which vendors appear in the technology stack. The question is who controls the recommendation process and what incentives shape it.

Vendor-neutral advisory starts with the client's problem and selects the appropriate technology. Vendor-captured advisory starts with a platform partnership and identifies problems the platform can address. The outputs can look identical from the outside. The difference shows up in the cases where the right answer is an open-source model, a different vendor, a smaller deployment than originally scoped, or no AI deployment at all.

In practice, vendor-neutral AI advisory looks like this: recommending open-source models in contexts where proprietary fine-tuning is not justified by the use case or the budget. Building data pipelines on portable infrastructure rather than vendor-specific APIs when portability is a legitimate business requirement. Evaluating two or three LLM providers against the client's actual performance criteria rather than defaulting to the incumbent partner. And, critically, being willing to recommend against deploying AI in areas where the ROI analysis does not support it.

That last point is the clearest test. An advisory firm with a platform dependency cannot easily recommend against deployment. Its commercial relationship depends on deployment happening. A genuinely independent firm can tell a client that a particular AI initiative does not meet the bar for investment. That recommendation, delivered clearly and early, is often the most valuable advice an enterprise receives in an AI engagement.

Redesign partners with technology vendors for infrastructure. The goal is to identify the right stack for each client's problem. That means the technology selection follows the problem definition, not the other way around.

The questions every enterprise should ask

Before engaging an AI advisory firm, ask these five questions.

1. Who does your AI advisory firm have commercial relationships with, and what do those relationships look like?

2. If their recommendation leads to Platform X, what is their financial relationship with Platform X?

3. Can they recommend not deploying AI, or recommend against a particular vendor, and still win the engagement?

4. Who owns your data pipelines and fine-tuned models when the engagement ends?

5. Is your AI strategy owned by your organization, or by your consultant's platform partner?

These are not hostile questions. They are the questions any competent procurement process should ask of any advisory partner. The fact that they feel pointed in the current market is itself a signal about how quickly the landscape has shifted.

What the acquisition actually validates

The OpenAI acquisition of Tomoro validates two things simultaneously, and it is worth being precise about both.

First, it validates that the enterprise AI consulting market is large enough and strategically important enough to warrant a $4 billion investment by the dominant AI platform. The demand for applied-AI advisory is real, substantial, and growing. Enterprises need help. They cannot navigate this technology landscape alone. That market reality is not going away.

Second, it validates that the independence of that advisory has become a scarcer and therefore more valuable asset than it was before. When the leading independent applied-AI consultancy exits the independent advisory market, the supply of credibly independent advisory contracts. The enterprises that want AI transformation without a platform agenda now have a clearer question to ask when evaluating advisory partners: are you independent, and can you demonstrate it?

The answer requires more than a disclosure. It requires an advisory process that visibly starts with the client's problem, a technology selection methodology that is transparent and multi-vendor, and a business model that does not depend on any single platform's adoption targets. Those firms exist. They are smaller than the newly consolidated players. But in a market where the value of independence has just increased, that is a meaningful position to occupy.