McKinsey is making the most branded bet in consulting: that the firm with a single, named architecture for AI will out-sell rivals who sell a grab-bag of tools. It calls that architecture the Agentic AI Mesh, and underneath it sits the deepest proprietary toolkit of any of the big three strategy firms, built by its data-science arm QuantumBlack and its Lilli assistant, with roughly twenty AI products and over a hundred reusable accelerators.

That shows up as a playbook called Rewired, sold to chief executives as the canonical way to turn AI ambition into measurable value, wrapped around three sticky numbers its partners repeat everywhere: a 20% profit uplift, a one-to-two-year payback, and three dollars back for every dollar spent. The firm runs the technology on itself first, and its leaders say AI has become part of the lifeblood of how McKinsey works. To deliver it, McKinsey leans on outcome-tied alliances with the cloud giants: an Amazon McKinsey Group aimed at billion-dollar transformations, a Google Transformation Group, and a Salesforce tie-up.

The honest part is that the mesh is mostly a way to organize advice, not a system the client runs alone. Its own partners concede most companies have not yet productized use cases, redesigned workflows around AI, or built the platforms and guardrails needed to run them at scale, and when a breach hit its Lilli environment in early 2026 the registry logged what happened but could not stop it. The framework is brilliant on the slide and thin on the live run-state.

Strip it down: no firm has a cleaner story or deeper proprietary stack. But McKinsey still bills the playbook through people, and the procedure walks out the door when the engagement ends.