What happens when a consulting firm puts a chatbot on top of a 35-year-old database and calls it AI transformation? Business publications run it as a workforce breakthrough.
McKinsey recently announced it had developed 25,000 AI experts within the firm. The actual story: they took an existing internal knowledge database, added a natural language interface - meaning employees can now ask it questions in plain English rather than running structured searches - and issued a press release. Major business outlets covered it without asking what "AI expert" meant in this context or what the underlying system actually was.
This is worth examining not because McKinsey is uniquely cynical, but because the same move is playing out across dozens of large organizations right now, and the people making hiring and investment decisions deserve to understand the pattern.
How the Template Works
Take something that already exists internally - a database, a knowledge management system, a process library. Wrap it in a large language model interface (LLMs are the AI systems powering tools like ChatGPT: they understand and generate human language). Issue a press release that emphasizes the AI layer while obscuring the existing asset underneath.
McKinsey has run versions of this playbook before. They helped sell ERP software (integrated business management systems) to corporations in the 1990s as complete organizational reinvention. Digital transformation in the 2000s reframed moving processes online as a strategic shift. "Big data" in the 2010s promised that collecting enough information would itself produce insight. Each wave generated real consulting revenue. Each wave also left companies with expensive implementations and results that were harder to measure than the original pitch.
AI in 2026 follows the same structure.
The Workforce Claim Deserves Scrutiny
The McKinsey announcement was framed as a workforce story: the firm created 25,000 experts. That language implies new capability, new skills, possibly new roles. What it actually describes is existing employees gaining access to a better internal search tool.
That's not worthless. Better search tools save real time. But it is not the same as developing 25,000 people with genuine AI skills - the ability to build, evaluate, fine-tune (adjust AI models for specific tasks and datasets), or deploy AI systems. The gap between those two things matters for anyone benchmarking their organization's AI progress against what large firms are publicly claiming.
Companies actually realizing productivity gains from AI tend to be specific and boring about it. They track ticket resolution times, defect rates in code review, time spent on first-draft document work. They measure outputs, not headcounts of "experts."
The noisiest AI announcements are usually the ones doing the least new work underneath.