OpenAI’s reported expansion of consultants, deployment managers, account directors, and solutions architects reveals a limit that model benchmarks cannot solve. Enterprise AI can sell itself in a demonstration, but production requires integration, workflow redesign, privacy controls, reliability, and organisational change.
The company reached an estimated $20 billion in annualised revenue during 2025, up from $6 billion in 2024, and now serves more than one million organisations. It is pursuing a $100 billion revenue target for 2027 while enterprise market competition intensifies.
Consulting is becoming part of the AI product because access alone does not create adoption.
The Production Gap Is Mostly Outside the Model
Industry figures cited by the source suggest 87 percent of large enterprises are implementing AI, while only 31 percent of use cases reach full production. Integration complexity, data privacy, and reliability remain leading obstacles.
These are architecture and operating-model problems. A strong model still needs authoritative data, identity, monitoring, evaluation, exception handling, and a workflow employees will actually use.
Consultants can help translate a broad capability into a bounded process with an owner and measurable outcome. They also support change management when automation alters authority, incentives, and the division of work.
The warning is that services can conceal product complexity. Enterprises should distinguish one-time adaptation to their business from permanent dependence on vendor specialists to keep the system functioning.
Rivals Choose Different Routes to the Customer
Anthropic has emphasised partnerships with Deloitte, Cognizant, and Snowflake, effectively using established firms as the consulting layer. Microsoft relies on existing enterprise relationships, Google bundles AI into Workspace and Cloud, and Amazon centres its strategy on AWS infrastructure.
OpenAI appears to be investing more heavily in direct engagement. That can place technical experts closer to customer problems and create faster feedback for product development.
It is also expensive and difficult to scale. Domain knowledge differs across banking, healthcare, manufacturing, government, and retail. A consultant who understands deployment mechanics may not understand the regulation or process being changed.
OpenAI’s reported enterprise share fell from 50 to 34 percent while Anthropic rose from 12 to 24 percent in foundation models. Implementation support may therefore become a competitive differentiator alongside model quality and price.
Enterprises Must Absorb, Not Rent, Capability
Vendor expertise can accelerate a first deployment, but customers need knowledge transfer. Internal teams should own data definitions, evaluations, workflow logic, incident procedures, and the business case.
Contracts should define deliverables beyond a working pilot: documentation, training, portable configurations, performance baselines, and a transition plan for ongoing operation.
The human challenge may be harder than the technical one. Forty-two percent of C-suite leaders reportedly say AI adoption is creating power struggles, conflict, and silos. Consultants cannot resolve those tensions unless executives clarify ownership and make decisions about changed work.
Organisations should also resist the sales race’s pace. A provider’s revenue target is not the customer’s transformation schedule. Deployment should expand when evidence, controls, and workforce readiness justify it.
This changes how buyers should evaluate an engagement. The strongest proposal will identify the workflow owner, expected baseline, required data, control design, employee training, and the point at which the customer’s own team takes over. A vague promise of transformation may create activity; a defined transfer of capability is more likely to create lasting operating change.
OpenAI’s hiring suggests the AI market is maturing into enterprise software, where implementation and customer success determine whether licences become durable revenue. It also confirms that the gap between capability and value remains large.
The winners may not be the vendors with the most impressive model in isolation. They will be those that help customers build systems they can eventually run without constant outside intervention. Consulting can bridge the production gap; genuine adoption is complete when the bridge no longer needs to carry the entire organisation.
