The Big Five want weeks of data plumbing before their AI runs, and the market is quietly walking away from it.

Every flagship consulting AI framework needs a data layer built first. That prerequisite used to read as diligence. Now it reads as a bill the buyer can refuse.

Start with the pattern almost no one names out loud. Every major consulting AI framework grounds its intelligence in a system that has to exist before the agent can run. McKinsey's approach needs pre-built domain ontologies and a semantic layer that defines the business across disparate systems. Deloitte's needs a single source of truth and a digital twin of the enterprise. Accenture's needs a data integration effort that typically takes four to twelve weeks. Wipro's needs a data modernization pass first. None of this is unreasonable as architecture. All of it is expensive as a sales proposition.

There was a moment when this was defensible. When frontier models were new and ran through a single cloud, the setup work was bundled with access, and paying for the plumbing felt like paying for the model. That moment has passed. Access is now everywhere, which means the data engineering no longer hides behind it. It stands alone on the invoice, and it is billed separately from the outcome it is supposed to enable.

Watch what the firms themselves are doing, because it contradicts the pitch. Wipro carved out a dedicated AI-Native Business and Platforms unit under Nagendra Bandaru, effective April 1, with its own operating model and an invest, build, partner mandate aimed at a services-as-software future. Cognizant agreed to buy Astreya for roughly 600 million dollars on April 29, acquiring managed-services muscle rather than building the integration layer itself, and paired it with a Project Leap restructuring expected to cut thousands of roles. EY put agentic AI in front of 130,000 auditors running on existing engagement workflows, not a freshly built digital twin.

Read the direction of travel. Each of those moves shortens the setup path rather than deepening it. When the firms selling the prerequisite are the ones quietly retiring it, the prerequisite was never the product. It was the meter running before the real work began.