A couple months ago, a great product manager I know got laid off.
Night before last, we talked in between games. Quick catch-up. Turned into a long conversation.
Thirty minutes in, he said something I couldn't stop thinking about.
"Everyone's talking about what AI can do. Nobody's talking about how any of this impacts the customer."
It was refreshing to hear and I agree 100%.
I've written a lot about this in a few different articles.
Yesterday I found research that explains why that observation matters more than most CEOs realize.
A 2024 Pega study found that 93% of business leaders claim to understand AI. Only 35% can accurately define it.
That's a 58-point knowledge gap.
But the execution gap is even worse.
Boston Consulting Group found that while 90% of companies pilot AI initiatives, only 10% scale successfully, an 80-point implementation gap.
Not a rounding error. A systemic failures costing companies millions.
After sitting with this data a while, I think I know why that gap exists.
And I think it's the same reason my friend is unemployed while his former company burns money on AI tools that won't work.
They're eliminating the foundation while buying the skyscraper.
Let me show you what I mean.
The Pattern Nobody's Naming (But Everyone's Living Through)
Here's what keeps showing up in the research.
Companies see headlines about AI replacing entire job functions.
They think... "We don't need as many engineers or product managers."
So they eliminate roles. Reallocate budget to AI platforms. Call it "efficiency."
The board applauds. Stock gets a bump. CFO looks smart.
But here's what the research shows happens next.
That 80-point implementation gap?
It exists because companies approach AI as a "procurement exercise" rather than a "transformation initiative."
Translation.
They're buying tools without building the foundation that makes tools work.
And what is that foundation?
It may not be what you think.
- It's not better data infrastructure.
- Not more computing power.
- Not the latest model.
It's the people who understand why customers buy.
What objections actually matter. Which feature requests to ignore.
How to distinguish between what customers say they want and what they actually need.
My friend wasn't just a product manager for that company.
He's institutional memory about product-customer fit.
The kind of knowledge that doesn't show up on a balance sheet. Until it's gone.
But by then it's too late.
Because you're 18 months behind competitors, wondering why your AI "transformation" feels like running in circles while standing still.
Here's what I think is actually happening.
When AI Sees Volume But Misses Meaning
AI can process a million customer interactions.
Your product manager handles maybe 50.
So the math seems simple: AI wins.
I get it. I'd think the same thing if volume was the only variable.
But that math misses something critical. And what it costs when you miss it.
AI processes 250,000 support tickets.
Reports: "Top issue: Login problems - 1,247 mentions."
Your engineering team spends six weeks rebuilding the entire authentication system.
New password flows. Biometric options. The works.
Cost: $120K in engineering time.
You announce it in your product update. "We heard you. Login is now 40% faster."
Three months later? Complaints haven't dropped.
Because your product manager would have caught it in the first ten tickets: "login problem" was customer shorthand for multiple different issues.
- Some couldn't access their dashboard after a password reset.
- Some were locked out after failed payment updates.
- Some hit errors when switching between accounts.
- Some were getting logged out mid-session on mobile.
The AI saw the word "login" and grouped them together.
Your PM would have seen...
"These are four separate product issues that customers describe using the same word."
You rebuilt the wrong system. Six weeks. $120K.
And the actual problem is still happening. Still costing you customers.
Here's the principle.
AI amplifies whatever foundation you give it.
Strong foundation, documented customer intelligence, proven patterns, institutional knowledge...
And AI becomes unstoppable.
Weak foundation, eliminated product managers, undocumented processes, no human curation...
And AI amplifies your blind spots at machine speed with minimal guardrails.
This is what the research means when it says only 11% of companies achieve "human-like AI conversations."
Not that AI isn't powerful.
McKinsey found that 43% of companies report AI productivity gains. But only 11% see measurable ROI at scale.
The 32-point gap between "it's working" and "it's profitable"?
That lives in the foundation you eliminated to buy the AI.
And here's the part that should make every CFO nervous.
By the time you realize this, the damage compounds faster than you can repair it.
The 36-Month Lag Nobody Saw Coming
Here's what the CFO's spreadsheet likely showed when they let my friend go.
- PM salary: $180,000
- AI platform: $50,000
- Net savings: $130,000
What actually happened.
The 18-Month Spiral
Quarter 1: $200K spent building wrong features (AI saw correlations, missed causation)
Quarter 2: Competitor launches what customers actually wanted
Quarter 3: Your ex-PM now works there, he knew from 50 conversations what your AI missed in 50M data points
Quarter 4: Emergency board meeting. Revenue down 15%. The consultant you hired for $400K delivers his diagnosis: "You're not 18 months behind. The gap is 36, they gained market position while you lost it."
You "saved" $130,000.
You lost $2,000,000. Minimum.
The spreadsheet showed the cost of having a product manager.
It didn't show the cost of NOT having one.
And look, if you're the CFO reading this, I get it.
Sometimes headcount decisions are out of your control.
Board pressure. Investor mandates. Market conditions force your hand.
This isn't about judging those decisions.
It's about seeing the downstream cost before you make them.
Because six months from now, when revenue is down and the board is asking why...
"We eliminated institutional knowledge before capturing it" won't be an acceptable answer.
Your job isn't to protect every role.
Your job is to extract the value from that role before it walks out the door.
That's the 90-day window most CFOs miss.
The one that turns a $130,000 savings into a $2,000,000 loss.
And maybe here's the most painful part.
While you spent 18 months learning this lesson through revenue loss, your competitors were doing something different.
They documented what their product managers knew before implementing AI.
If they were really paying attention, they enabled their product managers to spearhead that effort themselves.
- Which customer segments waste time.
- Which features look good in demos but fail in deployment.
- Why certain complaints signal churn risk while others don't.
- The "why we DIDN'T build this" decisions that save more money than the features you ship.
Then they trained AI on those proven patterns.
Same AI tools. Same starting position. Same market conditions.
Completely different outcomes.
One company eliminated institutional knowledge. The other activated it.
Your 90-Day Window Is Closing
And if you're a product manager reading this, here's what you need to know.
This isn't about your job being safe or not.
You and I both know the market doesn't care about fair.
This is about recognizing where things are headed and making a move before the decision is made for you.
Option 1: Lead the knowledge capture process now.
Document your institutional knowledge. Build the system. Show leadership exactly what institutional knowledge means in dollar terms.
You're not just protecting your role, you're demonstrating you understand the business at a strategic level.
Option 2: Document it for yourself.
Same process, different outcome.
You're prepared to walk out with a systematized version of your expertise. The patterns you've learned. The customer intelligence you've gathered.
That's not just a resume item. That's the foundation for your next role, your consulting practice, or your own company.
Either way, you're not waiting for the layoff announcement to figure out what your knowledge is worth.
You're capturing it while you still can.
The Sequence That Actually Works (If You Haven't Already Burned 18 Months)
The companies winning with AI aren't choosing between humans and machines.
They're asking a different question entirely.
"How do we capture institutional knowledge before we optimize anything?"
The 10% who succeed at AI follow this sequence.
Before you touch headcount: Document what your top product managers know that isn't written anywhere.
Start here, this week.
Schedule 90 minutes with your best PM. Ask them to walk through their last five major product decisions.
Record it.
The 3-Question Extraction Framework
Every product decision contains three layers
What data did you examine?
What data did you deliberately ignore?
Why? (This is where institutional knowledge lives)
Document 5 decisions = 15-25 extractable patterns
Time required: Three 90-minute sessions
Result: 80% of critical knowledge captured
Before you implement AI: Systematize what you just captured.
Most companies stop at "record the conversation." That's not enough.
You need to extract the decision framework.
"When we see X customer segment + Y behavior + Z complaint...we prioritize/deprioritize/investigate further because..."
Turn each of those 90-minute sessions into 5-10 documented decision patterns.
Simple format.
- Situation: [What was happening]
- Data considered: [What they looked at]
- Data ignored: [What they dismissed and why]
- Decision: [What they chose]
- Outcome: [What happened]
Before you scale: Test whether AI + framework can replicate good judgment.
Give a new PM a decision to make using only your documented systems.
Compare their conclusion to what your experienced PM would decide.
If the gap is huge? Your documentation is incomplete.
If the gap is small? You're ready to layer AI on top.
Then, and only then, you implement AI on that foundation.
This isn't about protecting jobs.
It's about protecting the knowledge that makes any job effective.
Because AI trained on your proven patterns beats AI trained on random data.
Every single time.
The 10% who succeed? They poured the foundation first.
The Honest Conversation About Who Benefits (And Why That's Not A Problem)
Now here's where it gets interesting, and where most companies get stuck.
This creates value for both the company AND the PM. And that tension?
It's not a bug. It's a feature.
The Mutual Interest Paradox
Company Fear: "Why document knowledge the PM can take?"
PM Fear: "Why document what makes me valuable?"
Both fears create the same outcome. Knowledge walks out undocumented.
The alternative?
Company Gets: Transferable institutional knowledge that survives transitions and enables AI
PM Gets: Systematized expertise that's valuable whether they stay or go
This isn't naivety. It's strategic alignment.
The PM who systematizes knowledge isn't making themselves replaceable, they're making themselves promotable.
The ability to extract and transfer institutional knowledge is rarer than the knowledge itself.
The company that enables this isn't being exploited, they're hedging against the 90% failure rate.
Both parties protect their interests. That's how the 10% who succeed think about this.
The alternative is worse for both of you.
If the PM leaves without documentation, the company loses everything AND the PM has nothing tangible to show for years of expertise.
Both parties lose.
If the PM hoards knowledge to protect their role, the company can't scale AND the PM becomes trapped in an undocumented job that's hard to explain to the next employer.
Both parties lose.
But when you document together, with mutual self-interest acknowledged, both parties win.
The company gets: Transferable institutional knowledge that survives transitions and makes AI implementation actually work.
The PM gets: A systematized version of their expertise that's valuable whether they stay or go. This isn't about trust. It's about aligned incentives.
The PM who leads this process isn't making themselves replaceable, they're making themselves promotable.
Because the person who can systematize institutional knowledge is more valuable than the person who hoards it.
And the company that enables this isn't being naive, they're being strategic.
Because institutional knowledge that's documented is better than institutional knowledge that walks out the door unnoticed.
Both parties protect their interests. Both parties win.
This isn't about trust. It's about aligned incentives.
The Question Nobody's Asking
My friend is still looking for work.
His former company is probably telling their board that "market conditions shifted."
That revenue is down because "the competitive landscape changed."
That customers are "more price-sensitive than forecasted."
Maybe all of that is true.
Or maybe they eliminated the people who could have told them what was coming six months before the board saw it in the numbers.
Meanwhile, their competitors...
The ones who documented and activated institutional knowledge before automating, aren't making excuses to their boards.
They're defending market position while others scramble to rebuild what they eliminated.
The question isn't "Can AI replace product managers?"
The question is "Can AI work without the institutional knowledge product managers hold?"
The research says no.
The 90% failure rate says no.
The 80-point execution gap says no.
Can you identify what institutional knowledge you have before it walks out the door?
Most companies discover the answer six months too late...
When revenue numbers reveal what their PMs could have told them at a fraction of the cost.
Your competitors are making one of two moves right now:
Move 1: Join the 90% who eliminate their customer intelligence infrastructure. Spend 18 months learning expensive lessons you may not recover from.
Move 2: Run a 90-day knowledge extraction. Capture the thousands of dollars hiding in institutional knowledge. Build the foundation that makes AI work.
There's no third option.
The window is getting smaller by the day.
The companies that document and activate these assets first won't just survive the AI transformation.
They'll own it.
Which move are you making?
Stay sharp,
Colin Taylor
Creator of The Asset Alchemy Method
P.S. If you know a company looking for an exceptional product manager in the Raleigh-Durham area, I'd be happy to connect you with my friend. He's the kind of PM who understands the difference between building features customers ask for, and solving problems they actually have. Reply to this message or DM me directly.
Sources
AI Knowledge Gap: Pega & Savanta Research Study (July 2024) "AI Challenges: Are Your AI Strategies Built on Sand?" cmswire.com
AI Implementation Gap: Boston Consulting Group (October 2024) "AI Adoption in 2024: 74% of Companies Struggle to Achieve and Scale Value" bcg.com
AI Value Realization: McKinsey & Company (January 2025) "The State of AI in 2025: Agents, Innovation, and Transformation" mckinsey.com
Frequently Asked Questions
What is the real cost of eliminating product managers before capturing their knowledge?
Research shows a pattern where companies save $130K in salary but lose $2M+ in downstream costs: wrong features built, competitors gaining market position, and an 18-36 month recovery timeline. The 80-point gap between AI pilots (90%) and successful scaling (10%) exists because companies treat AI as a procurement exercise rather than a transformation initiative built on institutional knowledge.
Why do 90% of AI implementations fail to scale?
According to Boston Consulting Group, most companies pilot AI without first documenting the institutional knowledge that makes those tools effective. AI amplifies whatever foundation you give it. Strong foundation with documented customer intelligence and proven patterns makes AI unstoppable. Weak foundation with eliminated expertise and undocumented processes means AI amplifies blind spots at machine speed. The Asset Alchemy Method addresses this by requiring knowledge extraction before any AI implementation.
How do you capture institutional knowledge before it walks out the door?
Start with the 3-Question Extraction Framework: For each major decision, document what data was examined, what data was deliberately ignored, and why. Schedule three 90-minute sessions with your best people. Document 5 decisions to extract 15-25 transferable patterns. This captures roughly 80% of critical knowledge and creates the foundation that makes AI implementation actually work.
Is documenting expertise risky for the employee who does it?
The opposite is true. The PM who systematizes knowledge isn't making themselves replaceable, they're making themselves promotable. The ability to extract and transfer institutional knowledge is rarer and more valuable than the knowledge itself. The Asset Alchemy Method shows that documentation creates aligned incentives where both the company and the individual benefit.