IBM's decision to train and certify tens of thousands of consultants on OpenAI's technology tells you less about models and more about distribution. The scarce input in enterprise AI is no longer the model; it is the trained human who can map a foundation model onto a bank's compliance stack or an insurer's legacy claims system. IBM is effectively buying a certified sales-and-delivery army before rivals lock up the same talent, and OpenAI gets a global services engine it would take years to build alone.
The global read is that enterprise AI is settling into a familiar shape: models commoditize, integration accrues the margin. As frontier capability converges and inference costs fall, differentiation shifts to the messy last mile of data governance, change management, and vertical workflow design. That favors incumbents with deep client relationships over pure model labs. It also creates a new dependency risk. When a services giant standardizes on one lab's stack and certifies its workforce accordingly, switching costs calcify. Clients who buy the certified capability inherit that lock-in whether or not they intended to bet on a single vendor.
For Japan, this is the most consequential storyline of the day, and it is largely being missed. The domestic IT market runs on the SIer model, where firms like the majors monetize headcount and long-lived maintenance contracts. An AI-certified-consultant channel is both a lifeline and a threat: it validates the services-led approach Japanese integrators know well, but it also imports the standards, tooling, and margins set in the US. If NT Data, Fujitsu, NEC, and the SIer tier do not build their own certified AI delivery capability fast, they risk becoming subcontractors executing playbooks written elsewhere.
There is a sharper edge for the RPA-heavy operations that many Japanese enterprises adopted over the past decade. Rule-based automation was a stopgap for labor shortages; agentic AI delivered through certified consultants can absorb much of that scope. SIers whose revenue leans on RPA seat licenses and maintenance should expect that base to erode, and should be repositioning those teams toward AI orchestration and governance now.
The practical move for Japanese decision-makers is to treat certification as strategy, not training budget. Fund internal AI delivery competency, negotiate multi-model flexibility into any partnership to avoid single-lab lock-in, and prioritize the data-readiness work that determines whether any of this delivers value. The firms that own the last mile in Japanese, in-context, and inside domestic compliance regimes will capture the margin. The rest will resell someone else's certification.
