HBR named it. MIT proved it. Nobody connected the two.
When asked what I think about AI in enterprise, I have to respond. What?
The one that doesn't pay for itself? Or the one that causes a productivity nightmare?
95% of enterprise AI investments are producing no measurable return. Tens of billions. MIT said it. The boardroom hasn't.
Workslop. HBR named it. Your people already knew it.
Content that looks like good work. Passes the visual inspection. Confidence without competence demonstrated.
Two hours per incident to clean up what the tool produced. Every incident. Every day. Across every department that handed the tool to someone without setting a standard for what the output needed to do.
Nobody has connected the two findings.
Yet.
And wait. There's more.
Three different data points. All pointing to one thing.
The research on why is already in. BCG's AI Readiness Report 2026 puts it plainly — 10% technology, 20% data and analytics, 70% people and process. Organizations that follow this outperform those that don't by 3x on ROI.
Most organizations are spending 90% of their attention on the 10% of the problem that matters least. The technology. The platform. The license. The tool.
That's not a technology problem. That's a knowledge gap wearing a strategy mask.
The difference between achieving a vision and creating a strategy uses the knowledge of what is possible and what is not yet possible with existing resources and abilities.
Some call it tactics. Vision without knowledge is hallucination. Strategy without knowledge is a mask. Tactics without knowledge is where the real damage happens — because tactics are where the money moves, where the decisions get made at speed, and where the errors compound before anyone notices.
That's what workslop looks like at organizational scale.
Traditional software came with a flat monthly fee. Use it or don't — the bill stayed the same. Agentic AI doesn't work that way. It bills like the electric company. The meter runs every time an employee opens a session, runs a query, chains a workflow, or asks it to check the weather. Nobody reads the meter until the bill arrives.
One unnamed enterprise spent $500 million on Claude AI in a single month. No caps. No governance. No checkpoint. By the time the invoice landed, it was too late to do anything but absorb it.
That's not an isolated case. Microsoft cut internal Claude Code licenses after engineers were generating between $500 and $2,000 in monthly AI costs per person. Uber burned through its entire 2026 AI budget by April. Amazon scrapped its internal AI usage leaderboard after employees were caught inflating token consumption to meet internal targets — carrying out needless tasks just to climb the league table.
Uber's CEO confirmed there is no link between AI tokenmaxxing and shipping useful products.
Same pattern. One critical difference.
Every previous wave gave you time. The damage arrived slowly enough that by the time it was undeniable, the people who made the decision had often moved on.
AI doesn't give you that grace period. The meter runs from day one. The workslop accumulates from the first session. The invoice arrives before the implementation is finished.
Same knowledge gap. Immediate impact.
In fact, you already knew this. Your instincts told you something was wrong before MIT named it, before HBR coined the word, before the invoice arrived.
You may have chosen not to listen. And that's ok.
Because that bill is a token of what it could be. The signal. Not the noise.
I've spent most of my professional career cleaning up other people's aisle 5's. That started with MRP. Then ERP. Then CRM. Then cloud. Same real problem in each instance.
The technology was capable. The organization wasn't prepared to get the most from it. The knowledge gap between the vision and what was not yet possible with existing resources produced the spill.
And someone called me in to clean it up.
Some might read this and think I oppose AI or its implementation. I don't. I use it every day for specific tasks where it earns its place.
What I know is this — companies call me in to clean up the spills on aisle 5. And it is a much better solution for preventing accidents before they happen.
Smart companies call me in before the spill.
And before you call anyone in — ask these questions. Independently. Anonymously. Everyone in your room. Collate the answers before anyone speaks.
What is the most important problem we are solving with AI — described in a way a five-year-old would understand it?
This is where vision starts. If your room can't answer this one with the same voice — stop.
You can't solve for what you don't know.
What is the real value of solving that problem?
Not the vendor's projection. Not the boardroom aspiration. The real value — to the customer, to the people doing the work, to the organization. If it can't be measured specifically, it isn't real enough.
What is the biggest gain we expect — and who feels it first?
Is it the customer? Then the people doing the work? Then the organization?
That's the sequence that works.
Can't be isolated for a proof of concept?
One problem. One team. One measurable outcome. If it can't be isolated, it's too big to test and too big to fail safely.
What results need to be achieved in order to move forward — or rethink the solution?
Define success before you start. Define the exit criteria before the meter runs. If you can't answer this before deployment, it may just cost you $500 million.
Five questions. Your room. Your answers. Your foundation.
This is the beginning of the conversation. Not the end of it.
Plan the foundation. Build it. Then hand them the tools.