OpenAI’s new AI consulting offering raises questions of trust, strategy

OpenAI on Monday unveiled a new company, the OpenAI Deployment Company, designed to help organizations build and deploy AI systems by embedding engineers specializing in frontier AI deployment, known as forward deployed engineers (FDEs), into their environments.

It features backing from 19 consulting and financial companies, and will include resources from OpenAI’s acquisition of AI consulting and engineering firm Tomoro, also announced on Monday.

Analysts and consultants said the move is in line with what various other AI vendors, including Anthropic, which entered the FDE space last week, have done or plan to do. But the OpenAI Deployment Company’s announcement revealed little about its ownership and control.

On one hand, OpenAI said that the new $4 billion company is “majority-owned and controlled by OpenAI.” But by not specifying the percentage of that ownership and the equity percentages owned by the various partners, it left open how much control and access will be in the hands of those backers/partners.

The announcement described the new entity as “a committed partnership between OpenAI and 19 leading global investment firms, consultancies, and system integrators,” but only named 15 companies, leaving the identities of the remaining four unknown.

Its statement also hinted that customers could get a market advantage, as its team might reveal and leverage not-yet-announced future OpenAI capabilities during engagements.

“The OpenAI Deployment Company FDEs will be able to build for where OpenAI’s frontier capabilities are headed, giving customers systems designed to improve as new models, tools, and deployment patterns come online,” OpenAI said. “Customers can move faster from day one, spend capital on durable systems and stay ahead of competitors by building around the capabilities that are coming next.”

The company did not reply to a CIO request for clarification or comment. 

Not like normal consultants

Ishraq Khan, CEO of coding productivity tool vendor Kodezi, said the new OpenAI company has to be evaluated differently, given the nature of FDE itself. 

FDE teams “are not like normal software vendors or consultants. To be effective, they need access to workflows, internal tools, business logic, data pipelines, permissions, and decision-making processes,” Khan said. “That means the question for CIOs is not simply ‘Does this model work?’ It becomes ‘Who is actually inside our operating system, what can they see, and who controls the people building around our most sensitive workflows?’”

The new entity’s complex ownership structure means that CIOs also need to be very careful about any agreements, Khan added. 

“Enterprises will need very clear contractual boundaries around data access, model training, employee access, subcontractors, audit rights, and where operational knowledge goes after the engagement ends,” he said. “Once engineers are embedded deeply enough [in an organization] to make AI useful, they are also close enough to create real risk if the access model is unclear. The next battle is who controls the deployment layer, because that is where enterprise value and enterprise risk both live.”

Added Justin Greis, CEO of consulting firm Acceligence: “This is a brand new organization with unprecedented levels of access. Who has access to whatever the teams learn?”

Multi-layered challenge

Frank Dickson, group VP for security at IDC, said he sees the new OpenAI company as making sense. But he sees the consulting partners, which include CapGemini, McKinsey and Bain & Company, less as investors than as deployment talent. 

The challenge of many enterprise AI deployments is multi-layered, he said. For some, it involves tweaking the underlying model, whereas others might need data cleanup and better tweaking of the data to work smoothly with the model.

But for many others, the problem is in all-but-infinite numbers of AI interactions throughout the enterprise environment. For that last kind of AI deployment problem, OpenAI’s skills may be less helpful, whereas the experience of a McKinsey, Bain or CapGemini may be critical.

Proceed with caution

But Sanchit Vir Gogia, chief analyst at Greyhound Research, urged CIOs to exercise extreme caution. That doesn’t mean that he suggests they resist what OpenAI is offering, but that they proceed cautiously, involving both legal to tighten contracts and the rest of IT to limit access.

“Majority ownership and control give buyers a headline answer, not a complete trust answer. The real diligence sits in board rights, reserved matters, information rights, partner access, FDE reporting lines, incident authority, data controls and customer contracting terms,” Gogia said.

“This changes the governance map. OpenAI brings the frontier model roadmap, Tomoro brings deployment talent, private equity brings portfolio scale, consultancies bring transformation muscle, and systems integrators bring implementation reach. That combination can accelerate enterprise AI, but it also introduces incentive complexity.”

Gogia added: “CIOs should not only ask who owns the shares. They should ask who governs the work. If OpenAI control is real, it must appear in the contract, the audit trail, the escalation process, and the data boundary. Ownership starts the diligence. It does not complete it.”

CIOs need a plain English governance disclosure before sensitive access is granted, he said. They should ask who approves FDE hiring, who sees deployment telemetry, who receives operational learnings, and who signs off on pattern reuse. “Trust often fails in the gap between headline control and contractual control. That gap is exactly where enterprises should look,” Gogia said.

The FDEs are privileged embedded AI deployment teams working inside the customer’s operating environment, he pointed out. That matters because FDEs redesign workflows, connect models to data and tools, build production systems and observe how work actually happens within an organization. “They see the formal architecture and the informal workaround. They see the process map and the exception queue. They see what the enterprise says it does and what it actually does on a difficult Tuesday afternoon,” he said.

Taken as a whole, OpenAI’s new venture “makes the FDE a privileged third party with embedded delivery authority,” he noted. “The exposure is intimate. They see process debt, manual workarounds, brittle APIs, compliance exceptions, data quality problems and the spreadsheet that refuses to die. The upside is speed from pilot to production. The danger is speed into opacity. CIOs should welcome capability, but govern proximity.”