The first AI-governance contract gets a price: ISG’s $17M deal.

For eighteen months, the agreed story in consulting was that the AI bottleneck had moved up the stack, from infrastructure to data, and from data to the work of running and governing what the models actually do. Hackett named it, PwC refined it, and Wipro gave it a governance frame. The thesis was right, but a thesis is not a product. As of this week, that gap closed, and governance now has a price tag.

The Hackett Group launched its AI World Class Benchmarks for the Agentic Enterprise on May 11, a packaged diagnostic that lets a company measure its AI transformation against outside benchmarks rather than a consultant’s judgment. Ted Fernandez put the frame plainly: AI is not technology-first, it is process-first, automation-specific design and orchestration. The bottleneck Fernandez and Hobbs had been naming for two quarters is now something you can buy, with a framework attached.

ISG moved in the same window. On its first-quarter earnings call, Michael Connors named a $17 million governance megacontract on the record, the first time a governance engagement has been publicly priced in this field. Twenty-one million dollars of ISG’s first-quarter revenue came from AI-related work, one-third of the total. The governance line is no longer a white paper. It has a contract value, and that is the threshold that matters. Wipro’s Ivana Bartoletti supplied the legal frame in an AI Magazine interview, introducing the principle of non-delegable responsibility: when a company deploys an autonomous agent, accountability for that agent cannot be pushed down the supply chain. That is not a thought-leadership pose, it is a liability framework, and someone will be paid to enforce it.

The consulting layer is packaging governance from one direction. IBM is occupying the same ground from another. At Think 2026 on May 5, Arvind Krishna made the most direct attack on consulting from any platform player in this field: AI is not helping your business, it is your business model. He paired it with watsonx Orchestrate repositioned as a vendor-neutral control plane for agents, not a deployment tool but the layer that governs how all the tools work together. The word neutral is carrying a lot of weight, because IBM is claiming the same governance layer above the models that the consulting firms are now productizing.

DeployCo entered from a third direction. The OpenAI forward-deployed consulting arm launched publicly on May 11 with $4 billion from 19 partners, including Bain, TPG, Brookfield, and SoftBank, and immediately acquired Tomoro and its roughly 150 forward-deployed engineers. Its stated job is enterprise AI deployment. Its unstated job is the same layer IBM and the consulting firms are racing to own, the governance and execution infrastructure that sits between the model and the business result. Matt Hobbs framed the stakes at the PwC and Google Gemini Enterprise launch: it is not a technology issue, it is an adoption issue, asking why a firm spending money and redirecting resources to drive an advancement is not seeing the returns it expected.

That formulation is now consensus across boutiques, the Big Four, and IT services. The question is no longer whether the bottleneck is real. The question is who gets paid to solve it. Hackett has a benchmark, ISG has a contract price, Wipro has a legal principle, IBM has a control plane, and DeployCo has capital and engineers. What the pure advisory firms, the ones whose governance story still lives in a slide deck, do not yet have is a packaged, auditable, repeatable deliverable. The firms that built the bottleneck thesis now have competitors selling the answer.