Everyone wants AI ROI. But let’s be honest — asking for it right now feels a little like asking an alpha product for quarterly user growth. Still, someone has to approve the spend. I get that.
But the number I keep hearing is 20%. Or 30%. Maybe 50% productivity gains.
I think the decimal is in the wrong place.
An anecdote and a build.
I’m at a point in my career where I want to work, and work hard. But my kids deserve real vacations — and laptops should stay home during them. Vacations with fishing, cabins, sundecks, maybe a jet ski or two.
So, with Q2 QBR landing squarely in the middle of summer, I set out to build something I’d never built before: a system to prepare my Quarterly Business Review before the quarter was even over.
The first key piece is the Visual QA layer. I wrote about this previously — the dashboard is no longer just the output; it becomes the place where we inspect and validate the work AI produces.
The second key point is that in preparing for this QBR through agentic exploration, I found myself diving deeper into the data and my understanding of what I needed to share than ever before. MUCH deeper.
QBRs are tough — lots of data, finite time between the end of the quarter and the meeting. Let’s call it two weeks — 80 hours. Realistically? 40, max. Analysis, slides, meetings, narratives, QA, all of it.
By the time I finished the skill-building for the prep work, what I had produced represented easily double or triple the amount of human effort I would normally have put into the work. And I did it all with roughly four hours of skill-building.
20%? You mean 1.2X? Pshaw. (Wordle joke.) No way. This was not a 1.2X improvement. It was a completely different category of output.
PLUS I never have to build the prep again. PLUS I can share it with colleagues.
It’s hard to know exactly what the ROI will be, and there is still plenty of debate about token efficiency (bring on the model routers!). But one thing I know for sure — it’s not 1.2X.
That decimal is in the wrong place.