Here’s a conversation I keep having with operators inside the Fortune 1000.
“How much value is your AI stack actually creating?”
A pause. Then: “We’re tracking it.” Then: “We’re working on the dashboard.” Then, eventually, an honest answer — “I have no idea, and neither does my CFO.”
These are leaders who can tell you, to the dollar, what every other vendor in their stack costs them. Salesforce ROI: they have a deck. Workday: they have a deck. Whatever DocuSign costs per signed contract: deck.
But the line item that doubled this year — AI — is the one nobody can defend in a board meeting.
This isn’t a tooling problem. The dashboards exist. The seats are paid for. The integrations are live. The problem is upstream: usage data lives in a hundred fragmented places, none of it joined to outcomes, and the people best positioned to read it (BU leaders, not central IT) are the last to see it.
So what gets reported instead? Anecdotes. “The marketing team loves it.” “Sales prospecting is faster.” “Legal saved two hours.” Maybe true. Maybe not. Definitely not auditable. Definitely not a board-ready answer.
We started Calyber because we kept watching this gap widen. Spend was going up. Confidence was going down. And the question — is any of this actually creating leverage? — was getting routed to the team least equipped to answer it.
Here’s what we believe.
AI usage data is operating data. It belongs in the same place you measure pipeline, throughput, and cost-to-serve. Not in a vendor dashboard. Not in a procurement spreadsheet.
The leverage isn’t in the model. It’s in the workflow it touched. Which means the question isn’t “are people using Copilot?” — it’s “what got faster, and for whom?”
Operators need a system of record before they need another tool. You can’t optimize what you can’t see, and you can’t see what’s smeared across seventy vendor portals.
This newsletter is the working notes of that thesis. What we see in the data when we connect it all up. What we hear from BU leaders when they finally get a clean read. What’s actually moving the needle, what’s not, and the patterns that show up across companies before they show up in the press.
A few things we won’t do.
We won’t run tool-of-the-week roundups. There are excellent newsletters for that — we recommend four of them on the site.
We won’t pitch the product in every issue. If you want a demo, we’ll get out of your way and book one.
We won’t pretend AI is going faster than it actually is inside large enterprises. Half of what’s published about “AI transformation” describes companies whose Copilot rollout hasn’t cleared procurement.
What you can expect.
One issue per week. Tuesday morning, US.
One specific observation per issue, grounded in usage data we’ve seen or patterns we’ve heard from operators across our customer base and network.
Occasional longer pieces when something deserves it.
If you’re a CIO, a head of AI or transformation, a BU leader inside the Fortune 1000, or an investor trying to figure out which AI bets compound — this is for you.
The next question after “how much value is our AI creating?” is “what would it take to answer that?”
We’re going to spend a few issues unpacking it.
— Tobias