I Told You So: McKinsey Just Proved the Value of Everything You Know
- Rich Washburn
- 25 minutes ago
- 5 min read
I don't do "I told you so" very often.
Actually, I actively dislike it. Technology moves too fast, predictions are too easy to cherry-pick, and there's already enough victory-lapping on LinkedIn to power a small country.
But screw it...I told you so.
And I'm not saying that for me. I'm saying it for every 50-something-year-old person who has been wondering whether artificial intelligence just made thirty years of accumulated experience obsolete.
Because McKinsey just handed you the receipt.
I wrote that consulting firms were sitting on an enormous problem. Their real product wasn't PowerPoint. It wasn't billable hours. It wasn't even consultants.
It was everything those consultants had learned.
The weird stuff. The undocumented stuff. The hallway-conversation stuff. The "we tried that in 2008 and here's why it blew up" stuff. The judgment that develops after you've been punched in the face by reality enough times that you stop confusing a framework with the world the framework is supposed to describe.
I called it tribal knowledge.
And I argued that if firms didn't start extracting it and turning it into machine-usable knowledge systems, they were going to lose it one retirement party at a time.
My exact warning was: codify or die.
McKinsey is now building 25,000 AI agents.
So... yeah. I told you so.
BUT THE IMPORTANT QUESTION ISN'T WHAT MCKINSEY DID.
The important question is: what did McKinsey have before AI arrived that allowed it to do this?
And the answer is decades of captured intellectual machinery.
That's where this gets interesting for you.
Because I've been making this argument in layers.
First, I argued that organizations were sitting on tribal knowledge that would disappear one retirement at a time unless they deliberately codified it into machine-usable systems. The real asset inside a consulting firm isn't labor — it's accumulated knowledge embedded in people, conversations, client war stories, intuition and undocumented operating patterns. The strategic mandate is to extract that knowledge, teach AI how the firm thinks, and scale insight instead of headcount.
This isn't about robots replacing consultants. It's about machines remembering what humans forget.
Then I argued that the practical method for extracting knowledge from a human being is simpler than people think. Talk. Record. Transcribe. Organize the old speeches, the stories, the scattered content. Casual conversation surfaces knowledge that structured processes often miss. AI can organize and amplify that material without being the source of the underlying expertise.
Then I pointed out that older experienced professionals are sitting on precisely the thing the AI economy is going to need. Build the column. Externalize what you know. Make your knowledge available to machines.
And now McKinsey: 25,000 agents.
That's not three disconnected articles.
That's a thesis playing out in real time.
YOU DON'T NEED MCKINSEY'S BILLIONS.
You don't need 25,000 agents. You don't need to train a foundation model.
You need a corpus.
And the method for building one is already documented. Talk. Record. Write. Publish. Explain your decisions. Capture failures. Save the stories. Document exceptions. Preserve your weird little shortcuts. Explain why you do something differently than everybody else.
That casual consulting method — recording conversations, capturing stories, transcribing material — isn't just a consulting methodology anymore when viewed through this lens.
It's a knowledge-extraction protocol.
Human experience. Natural conversation. Captured corpus. Structured knowledge. AI-accessible capability.
McKinsey just did the enterprise version.
And the argument to the 50+ crowd becomes almost painfully simple.
MCKINSEY HAD MCKINSEY TO ENCODE. WHAT DO YOU HAVE TO ENCODE?
Because some 58-year-old plant manager has 35 years of operational knowledge nobody else possesses. Some 63-year-old CPA knows exactly where deals go sideways. Some 55-year-old nurse knows things about patient behavior that will never appear in the clinical manual. Some 60-year-old electrician can hear something wrong with a motor before anybody's diagnostic equipment notices it. Some 57-year-old salesperson can tell within three minutes whether the buyer sitting across the table actually has authority.
That's data now.
Not Big Data. Not telemetry. Not another corporate data lake full of clickstreams and invoices.
Human operational data.
And almost nobody has bothered capturing it.
That's the vacuum.
And McKinsey isn't proof that the vacuum disappeared. McKinsey is proof that the people who started filling it early now have an extraordinary asset.
That's much stronger than "AI is replacing consultants."
Your thesis is almost the inverse: AI makes accumulated human expertise scalable for the first time.
And therefore the older person with legitimate expertise shouldn't be panicking about AI. They should be furiously documenting themselves.
Because the 25-year-old has AI. Great. Soon everybody has AI.
But the 25-year-old doesn't have your thirty years.
Unless you leave those thirty years lying around undocumented long enough for somebody else to reconstruct them.
AND YES, ANALYSIS JUST GOT CHEAP.
Much of what consulting firms historically charged enormous amounts of money to produce is becoming commoditized. Market analysis. Benchmarking. Research synthesis. Competitive matrices. First-pass strategic options. Financial models. Presentation structures. Executive summaries.
Machines are getting very good at that layer.
Which means value moves. Not away from humans. Toward harder human problems.
Judgment. Accountability. Relationships. Politics. Execution. Leadership. Changing behavior. Getting seven departments with seven different incentives to agree on something. Knowing that the CEO saying "yes" in the meeting absolutely didn't mean yes. Knowing which person needs to be brought in privately before the next meeting or the project dies.
AI doesn't eliminate any of that. If anything, once everybody has inexpensive analysis, those abilities become more visible.
The PowerPoint was never the hard part. Getting Monday morning to operate differently from Friday afternoon was.
THE MODEL ISN'T THE MOAT.
The value isn't increasingly concentrated in the foundation model. It's moving into the systems surrounding it. The harness. The workflow. The data. The institutional memory. The interfaces. The operating environment.
Claude doesn't become McKinsey because you open Claude. GPT doesn't become McKinsey because you type "act like a McKinsey consultant." That's cosplay.
The interesting system emerges when you combine capable models with decades of accumulated methodology, proprietary information, structured workflows, evaluation, context and institutional knowledge.
Only an organization the size of McKinsey could once afford to build that. Today, you can build a small version around yourself. Not 25,000 agents. Maybe one. But that one can contain an awful lot of you. And it keeps getting better.
SO YES. I TOLD YOU SO.
McKinsey spent decades building McKinsey before AI showed up. That's why McKinsey can now build an AI McKinsey.
You don't have decades left to start documenting what you know.
Fortunately, you don't need them.
You have a phone. You have a microphone. You have a lifetime of scar tissue. And you now have machines capable of turning that scar tissue into leverage.
Start talking.
Build the column.
Because McKinsey just showed you what happens when the AI finally arrives and your knowledge is already waiting for it.
I told you so.
Rich Washburn is a technologist and strategist working at the intersection of AI, infrastructure, and capital. He is Managing Partner and Chief AI Officer at Eliakim Capital.