AI agents are killing consulting

Welcome back to Building AI Agents, your biweekly guide to everything new in the field of agentic AI!

How are you supposed to feed those babies and buy a house in 2025 if your critical business workflows aren’t automated??

We’re looking to make Building AI Agents even better, and we’d love it if you could help us by answering the poll question below News!

In today’s issue…

• How agents are wrecking Big Consulting’s business model

• OpenAI’s new browser will be run by ChatGPT agent

• Great use-cases for AI agents—and terrible ones

• 10 essentials for enterprise MCP servers

• The best agents from OpenAI’s GPT-5 hackathon

…and more

🔍 SPOTLIGHT

Some advice for big consulting firms: learn to leverage AI agents, or you’re toast. That consultation will be $10 million, please.

A new Wall Street Journal report (paywalled, but there’s a decent open summary here), lays out the “existential” shift occurring in consulting giant McKinsey as agentic AI rapidly automates much of the work that it traditionally does. Already, the company has reduced its headcount by over 10%.

Why is the agent revolution hitting consulting particularly hard? As I laid out in the July 28 Spotlight, the tasks that are most liable to be automated by agents are ones that constitute major pain points for many businesses and involve interpreting and creating large amounts of text (remember, agents are powered by large language models). Much of what consultants do is exactly this: do a deep dive on a client and its industry, and make recommendations for strategic improvements.

If you’ve ever tried ChatGPT’s Deep Research functionality, you’ll know that churning through hundreds of pages of text and synthesizing them into a concise report and set of recommendations is something that LLMs were born to do. Of course, LLMs are fallible—but so are consultants. When the price difference between the two is cents for the former vs hundreds of dollars an hour for the latter, it’s no wonder that the industry sees agents as an existential threat.

McKinsey hasn’t been taking this lying down. At the same time that it’s been cutting jobs, it’s rolled out around 12,000 agents. Projects that previously would have required a whole team, including a partner, now use just 2-3 consultants, augmented by deep research agents. 40% of the firm’s revenue now comes from advising on AI and related technology.

So where does this leave agent builders? Well, remember my prescription for finding agentic automation opportunities: text-heavy workflows that constitute major pain points for many businesses. McKinsey and its main rivals, Boston Consulting Group and Bain & Company have a combined annual revenue of around $36 billion. Clearly, management and strategy consulting is in demand.

Only the biggest companies, though, could afford these kind of services—until now. Armed with AI agents, independent consultants targeted towards small and midsize businesses could soon have their day in the sun. Picture a solopreneur with some experience in an industry—say, restaurants—who sells consulting services to local eateries for a few thousand dollars, not the millions that Big Three firms charge their clients. A couple rounds of transcribed interviews with a restaurant’s management fed into LLMs, some work on its books by data analyst agents, and a deep research agent to synthesize it all into a set of recommendations, and you’re in business.

There are countless opportunities just like this being created by the rise of agentic AI. All they require are enterprising people to seize them.