Was the AI promise kept?
Your AI works. That is not the problem.
That is the problem.
Bain surveyed 951 companies this year. Their conclusion, in their words: "The technology worked. The value didn't arrive."
That is not my line. I wish it were. It is the cleanest description of the enterprise AI gap anyone has written, and it comes out of Bain's Automation and AI Pathfinder Survey 2026.
I have seen that exact gap before. I saw it inside an electric company, standing in a room full of radar screens.
Late 1980s into the 90s. I was consulting for Saturn Technical Services, Brian Fitch's firm. Commonwealth Edison — the electric utility for all of Chicagoland — was moving its business unit out of downtown Chicago to Oak Brook Terrace.
My contract was the data migration. Wang/VS to Novell NetWare. Move every business user onto the new environment without the 24-hour call center ever going dark.
That was the job. The education was everything I saw while doing it.
ComEd had Doppler radar feeds. When a storm came into the service area, the operations team did not sit and wait for the phone to ring.
They watched the radar. They tracked the storm path. And they drove the repair trucks out and parked them along the route the storm was about to take.
Then the storm hit. Lines came down. Outage calls started coming in.
And the trucks were already there.
Not dispatched. Already there. Crews worked from one downed line to the next in the same corridor the storm had just crossed. Customers got power back in hours instead of days.
Every other utility in that era had the same weather data. It was on the news. Anyone could see it.
They waited for the phone to ring, then sent the truck.
Same technology. Completely different business.
Here is what nobody wants to hear.
ComEd's advantage was not the radar. The radar was a commodity. You could watch a storm come in from your living room.
The advantage was that somebody had rebuilt the dispatch workflow around the radar. Somebody decided crews would roll before the calls came in. Somebody rewrote what a shift looked like. Somebody had to defend that to a budget committee — paying crews to sit in a parking lot in Aurora waiting for a storm that might turn.
That is not a technology purchase. That is a workflow redesign, with a name attached to it.
The utilities that stayed reactive were not underinvested. They had the same feeds. They just plugged new technology into an old process and expected a new result.
They bought the radar. They kept the phone.
Read that line again, then go look at what your company did with AI last year.
The 2026 numbers are ugly, and they are not mine — they belong to Bain.
From Bain's 2026 B2B Growth Agenda (a survey of more than 1,100 senior executives across 18 industries, published March 2026):
Same executives. Same confidence. Bigger number.
From Bain's Automation and AI Pathfinder Survey 2026 (951 companies, published June 2026):
That 44% is not optimism. It is a circular bet. It is buying a new truck with money you expected from a truck that never showed up.
I have watched this movie. I saw what happened when the government broke up AT&T. I survived International Harvester collapsing into Navistar. I watched MCI/WorldCom evaporate. The confidence right before a correction always looks identical to the confidence during the boom. That is what makes it a correction.
Meanwhile the companies at the top — the ones Bain found had embedded AI into redesigned commercial workflows, end to end, with clear ownership — are reporting roughly 2x the AI-driven revenue growth and 1.8x the cost efficiency of their peers.
Same technology. Completely different business.
Just like the trucks.
Three reasons. All three are ComEd reasons.
One: you optimized a step instead of rebuilding the process. You put a copilot inside a workflow that was designed for humans doing it the old way. The step got faster. The process did not change. Nobody at ComEd made the call center faster. They made the call unnecessary.
Two: nobody owns the outcome. You have an AI lead. That is not the same thing. An AI lead owns deployment. Somebody has to own the number. At ComEd, somebody had to defend paying crews to wait in a parking lot. That person owned the outcome, and they were accountable for it. Your AI pilot has a dashboard, not a defender.
Three: you counted the license, not the friction. The best technology decision I saw at ComEd was not the radar. It was the handheld meter reader. Before it, a meter reader walked onto a customer's property — the yard, the side of the house, sometimes the basement — to read the box. The handheld let them pull the reading from the sidewalk.
The customer's day got easier. The meter reader's job got easier. The company saved labor hours. And the data got cleaner, because readings happened on schedule instead of waiting for someone to be home.
One decision. Friction removed for three parties at once.
That is what value capture looks like. Not a faster step. Friction gone, across every stakeholder, at the same time.
Ask yourself honestly: does your AI deployment do that? Or did it just make one team's step 20% faster while everyone else's day stayed exactly the same?
Not a transformation program. A reckoning. Four moves.
Not because the technology fails. Because they inserted AI into processes built for the old way of working. The step gets faster. The process — and the economics — stay the same. Bain's phrasing: the technology worked, the value didn't arrive.
Bain found that companies embedding AI into end-to-end redesigned workflows, with clear ownership, report roughly 2x the AI-driven revenue growth and 1.8x the cost efficiency of their peers.
Bain's Automation and AI Pathfinder Survey 2026 found 44% of companies — the largest group — plan to fund generative and agentic AI investments from savings generated by prior automation programs. For many, those savings never materialized.
Only 7%, per Bain's 2026 survey of 951 companies. Most systems still route decisions to a human — while the investment case assumed full automation.
No. The models work. The demos are real. The failure is organizational: no data foundation, no workflow redesign, no owner of the number.
Compare projected value to captured value in dollars. If you can only report usage and adoption, you do not have an answer. You have a dashboard.
The measurement takes days. The workflow redesign takes a quarter. The willingness to admit the first wave underdelivered takes the most courage and the least time.
The storm is coming either way.
Everybody in your market can see it on the radar. The data is not the advantage. It never was.
The only question is whether your trucks are already parked on the route — or whether you are sitting by the phone, waiting for it to ring, with the best AI stack money can buy running quietly in the background, doing nothing that anyone can measure.
I work with four executives at a time as a Fractional CDO. Ninety-day minimum. We find the phone you kept, and we take it off the wall.
M.A.D. (Maverick Advantage Design) Designs Your Brand. M.A.P. (Maverick Advantage Platform) Makes You Known For It.
Stop Reading. Start Seeing.
— Charles K. Davis
Fractional CDO
seriodesignfx.com
P.S. If your AI report to the board leads with adoption rates, you already know the value number is bad. You just have not been asked for it yet. You will be.
