For decades, consulting firms made their money selling advice — packaged in slide decks and billed by the hour.
Now, developers are using AI to mimic that process, making knockoff McKinsey consultants available right in your browser.
Take Vercel’s new “skills” library, an open-source repository of nearly 90,000 reusable skills for AI agents. They range from copywriting and code review to consultant-style problem solving. Vercel, an AI startup valued at over $9 billion, runs a cloud-based AI platform for developers.
“Skills,” in the AI sense of the word, are capabilities that developers can create or download and give to an AI model or agent so it can perform a specific task without having to train the model from scratch.
The idea gained traction after Anthropic introduced “skills” for its chatbot Claude in October, which helped popularize the concept. Since then, developers have begun building and sharing skills that can be plugged into various AI systems.
Business Insider reviewed Vercel’s skills library and found at least four labeled with the term “mckinsey” and 26 skills labeled with the term “consultant.”
The most popular consulting-related skill at Vercel is labeled “mckinsey-consultant.” It was first uploaded on January 25 and is so far averaging 445 installs a week. That’s a respectable number, but still leagues from the most popular agents in Vercel’s library, which can have hundreds of thousands of installs.
It also has 200 stars on GitHub, which means it’s pretty popular, and has passed through a couple of security audits — a sign that it’s both viable and gaining traction among developers. In short, users are finding it useful.
Vercel’s library describes the skill as a prompt framework — originally made for Claude — that guides AI through defining problems, generating hypotheses, conducting structured analysis, and creating slides, replicating the classic workflow of a typical McKinsey consultant.
Business Insider asked a former McKinsey staffer to examine the McKinsey-style agent uploaded to Vercel’s library and see how it stood up against the real thing. Arvind Vasudevan, a former McKinsey engagement manager, told Business Insider that it lacks a key ability that defines a McKinsey consultant.
“It misses the point of how MBB and strategy consultants add value,” he said in a text message, referring to the group of big consulting firms that include McKinsey, BCG, and Bain. “A large part of the value is the questions consultants ask + conversations they have that help clarify thinking, uncover unstated assumptions, and ensure deep thinking. None of that is happening in this agent, which is doing a set of boilerplate analysis without that Socratic questioning and thinking.”
They may not be the real thing, but AI agents that mimic the work of consultants are already driving millions in revenue for companies like PromptQL, an AI enterprise platform launched by open-source unicorn Hasura.
The platform helps clients build custom AI analysts by integrating their internal data with the foundation models they already use. Once deployed, these AI analysts can perform tasks typically handled by data scientists or engineers — and continuously learn and adapt to their environments over time.
PromptQL’s cofounder and CEO, Tanmai Gopal, told Business Insider that the biggest barrier — or moat — to selling analysis is understanding the relationships between people, data, and revenue.
“McKinsey’s teams spend weeks embedded inside a company absorbing how it actually operates: the exceptions, the tribal knowledge, the definitions that differ between departments. That company-specific context is what makes their advice worth millions,” Gopal told Business Insider.
Gopal said enterprise AI tools often fail because they lack proper grounding, tending to guess rather than ask questions, learn from feedback, or maintain shared understanding across teams.
PromptQL, he said, aims to address those issues through a shared layer of understanding that adjusts with every new input.
“When a team member corrects the AI, teaches it a definition, or resolves an ambiguity, that knowledge becomes permanent and available to everyone. It’s not a semantic layer that data engineers maintain. It emerges from conversations,” he said.
Models don’t automatically know internal nuances — like pricing changes, team-specific terminology, or conflicting definitions of revenue. The real issue isn’t capability, but missing context, Gopal added.
In other words, the consultant’s slide deck was never really the product. It’s their judgment — and that’s the part AI is still learning.
Something to share about how consultants are using AI? Business Insider would like to hear from you. Email Lakshmi Varanasi at [email protected] or contact her on Signal at lvaranasi.70.
