Bookings Are Up. The 6% Hasn't Moved.

Notice what makes this credible: the target is split fifty-fifty between revenue and cost. Most AI value claims live on only one side of the ledger — a productivity anecdote, or a savings estimate — because that's easier to produce than proof of both. A number that has to show up in two places is a better test of whether an AI program is real.

Short post: "The AI number that has to show up twice." Use the revenue/cost split as a simple credibility filter executives can apply to any AI ROI claim they hear this quarter — including their own team's.

Two weeks ago, the missing budget line in most AI business cases was inference cost. This week it's identity and access. Every agent you stand up is quietly minting credentials that someone has to own, and most companies can't currently answer "how many AI agents are live right now, and who governs their access."

Ask-the-room exercise for a leadership offsite: how many AI agents does your organization currently have running, and who owns the identity list for them? Most rooms don't have an answer — which is itself the point.

It would be easy to read soaring AI bookings as proof the ROI question is settled. It isn't. Bookings measure what enterprises are willing to pay for — not what they're getting back. The gap between "we bought it" and "it moved earnings" is exactly the gap this brief exists to track, and this week that gap got wider, not narrower.

One visual: a steep vendor-revenue growth curve next to a flat earnings-impact line over the same period. The image makes the argument better than any paragraph — worth turning into a single-graphic LinkedIn post.

This is the AI-and-headcount debate playing out in public in real time, with the employer and the union disagreeing on the cause in the press. What's notably absent from both sides: a dollar figure. That absence is the tell — this is a narrative dispute, not a documented cost case, though it will likely get cited as one.

Practical piece for any executive planning a visible AI rollout: decide how you'll explain workforce changes before you deploy, not after. Use this case as the cautionary example of what happens when that sequencing gets skipped.

This is what unpriced AI IP exposure looks like when it actually goes wrong — two major companies accusing each other of destroying evidence over what moved with an employee. Almost no mid-market company has an AI-specific clause in its vendor or employment agreements covering this exact scenario. The lawsuit is the argument for writing one before you need it.

Practical, not newsy: three questions every executive should be asking their own counsel this month about AI vendor contracts and employee mobility. The headline case is the hook; the value is the checklist.

The most reliable AI ROI motion we've run this year isn't a pitch — it's proof, handed over first. It converts because it isn't a claim about what AI could do for someone; it's a demonstration on their own business, dated and specific. That's a pattern any executive readiness program worth its fee should be able to point to in its own operations, not just in a client's.

Walk the pattern end to end without naming anyone: the discovery questions that reveal what someone's already straining to keep up with, the personalized brief that comes back, the conversation that follows. The value of the piece is that it's reproducible by the reader, not just admirable.