McKinsey Wants to Put AI Agents in Your Back Office. It Also Owes AppliedAI.
McKinsey wants to put AI agents in the back office. The efficiency numbers it is citing to prove the case come from the company McKinsey is simultaneously telling clients to use.
McKinsey and AppliedAI announced a partnership on May 22 designed to help enterprises deploy AI agents in regulated workflows. The pitch is immediate: McKinsey published research the same day showing 62 percent of enterprises are experimenting with AI agents but only 23 percent have scaled one to production. That gap is the problem the partnership promises to solve. A leading European chemicals manufacturer compressed a two-week vendor onboarding process to under five minutes of active processing, according to AppliedAI and McKinsey's press release on PR Newswire, which the companies say represents a 99 percent-plus reduction in manual effort.
There is a tension buried in the announcement. McKinsey's press release states that "McKinsey has a financial interest linked to the performance of AppliedAI." McKinsey is simultaneously the advisor pointing out the enterprise adoption gap and a stakeholder in one of the firms positioned to close it.
The efficiency figures are self-reported. The chemicals manufacturer is unnamed. No independent metrics have been published.
The 62 percent and 23 percent figures come from a McKinsey survey conducted in November 2025. The survey methodology — sample size, response rate, whether respondents were primarily McKinsey clients — is not disclosed in the available materials. McKinsey's press release cites the numbers without characterizing them as self-reported enterprise claims rather than independently audited performance data.
AppliedAI is an Abu Dhabi-based AI company with over 350 employees and deployments across the United States, Europe, and the Middle East, according to Wamda. The company raised $55 million in Series A funding in February 2025 at a $300 million pre-money valuation, Wamda reported.
Consulting firms have sold enterprise transformation before. In the 1990s, business process reengineering carried the same basic promise: redesign how work flows, eliminate manual overhead, compress timelines. The engagements that generated the most documented post-mortems — including AT&T's massive 1994 reengineering failure and Ford's accounts payable redesign, widely cited in subsequent academic analyses — showed a consistent pattern: the firms that advised on the transformation also sold the software implementations and benefited from longer engagement cycles regardless of whether the promised outcomes materialized. The structural parallel to today is well-documented in the management literature: the incentive to overpromise efficiency gains is baked into the business model when the same firm is both advisor and vendor beneficiary.
The language has changed; the structure has not. McKinsey's current materials describe the partnership as designed to "rapidly rewire regulated enterprise processes with AI" — phrasing that echoes the 1990s era without needing to reach back for a specific engagement. Whether the current agentic-AI efficiency claims will prove more durable than their BPR predecessors is an open question that independent industry analysts have not resolved: the survey McKinsey cites does not show audited outcomes, only stated intentions.
Regulated enterprises running procurement, compliance, and vendor management through AI agents are precisely the workflows that SAP, Oracle, and ServiceNow built their businesses around. A successful Opus deployment doesn't just modernize a process — it displaces the incumbent software layer that currently manages it. The consulting firm positioning itself as the advisor also holds a financial stake in the vendor most likely to disintermediate the enterprise software incumbents. That creates a conflict of interest that runs in more than one direction.
If the efficiency claims hold at scale, the economic case is real. If they rest on a single unnamed client and a consultant with a financial stake in the vendor, the buyer should ask for the arithmetic.
Before signing a contract with a consulting firm that also holds equity in a recommended vendor, buyers and auditors should ask two specific questions: What is the size of your equity interest in AppliedAI, and do you have a formal conflict-of-interest disclosure policy for vendor recommendations that apply to this engagement?