Within a roughly two-week window in May 2026, at least three of the four largest professional-services firms standardized their delivery organizations on a single AI model vendor, which makes the model an enterprise's auditor and implementation partner have adopted an input to that enterprise's own architecture decisions and a concentration risk that vendor-neutral strategy advice does not surface.

Anchored on May 2026 standardization moves: PwC and KPMG each committed their delivery organizations to Anthropic's Claude within roughly two weeks of each other; Deloitte has built on Claude since late 2025; EY built its program on Microsoft and OpenAI. VERIFIED 2026-05-29 via the Anthropic newsroom: PwC 30,000 certified (14 May, anthropic.com/news/pwc-expanded-partnership), KPMG 276,000-employee rollout (19 May, anthropic.com/news/anthropic-kpmg), Deloitte expanded alliance late 2025 (anthropic.com/news/deloitte-anthropic-partnership); these figures are now stated in the article body. EY built on Microsoft Azure plus OpenAI (its EY.ai platform was a $1.4B EY investment in 2023; a SEPARATE $1B Microsoft-EY client co-investment was announced May 2026 and must not be conflated with EY's own program). Production-model disclosure is carried in the body: this publication is written by Claude (Anthropic), and three of the firms standardized on Claude, so the analysis is buyer-side and symmetric by construction (it would read identically for any house model, and EY's Microsoft/OpenAI bet carries the same concentration question). Scope: an observation about vendor concentration in the advisory channel and its architectural spillover, NOT a quality ranking of any model and NOT a claim that any firm chose badly. Sources verified 2026-05-29 (anthropic.com/news/pwc-expanded-partnership, anthropic.com/news/anthropic-kpmg, anthropic.com/news/deloitte-anthropic-partnership; EY stack via the EY newsroom and CIO Dive). 90-day review cadence (27 Aug 2026). Trigger conditions to revisit before next cadence: (a) the firms publicly maintain genuine multi-model delivery with no downstream pressure on client architecture, which would move the claim toward Partial; (b) a reversal or a fourth-firm move materially changes the concentration picture; (c) documented evidence that house-model standardization did or did not steer client reference architectures. Related: AM-185 (/frontier-labs-as-systems-integrators/, drafted and awaiting publish-move, the labs moving down-stack into integration, the same concentration question one actor upstream) and the published agent-protocol-tax piece (/agent-protocol-tax-mcp-a2a-llama-stack/, the lock-in question one layer down in the technology stack).