Why Did One AI Pilot Cost $310,000 More to Kill Than the Other?
In a composite case outlined by Dr. Jonah Tebaa, one AI pilot cost $310,000 more to kill because the sponsor who built its funding case occupied the review seat, allowing it to draw budget for five months after missing an unblock-rate target of 15 percent. Conversely, the cheaper pilot was shut down within two weeks because an independent reviewer enforced the numeric stop rule on schedule, exemplifying Dr. Tebaa's framework requiring non-sponsor reviewers.
Dr. Jonah Tebaa opens with two composite AI pilots inside the same six-initiative, $2.4 million portfolio, both missing an identical unblock-rate target in the same review cycle. One pilot was killed within two weeks. The other kept drawing budget for five more months and cost $310,000 more before it was shut down anyway, still short of its own target.
Dr. Tebaa is explicit that the case is composite and illustrative rather than drawn from a single client's file, and that every dollar figure in it should be read that way. The pattern, he argues, is what recurs across the portfolio reviews he has observed - not one company's numbers.
The Same Number, Two Different Reviewers
Both pilots, in Dr. Tebaa's account, had a stop rule that was numeric and specific before launch - an unblock rate below 15 percent by month four. Neither failure, he argues, traces back to a vaguely written criterion; both organizations had done that part of the work correctly. What separated the two outcomes was who occupied the review seat when the number came up short. In the costly pilot, the sponsor who had built the funding case was also the person deciding whether to call it a failure. In the cheaper one, an independent reviewer with no stake in that initiative's outcome held the seat instead.
Four Fixes Before the First Dollar Moves
Dr. Tebaa's framework asks organizations to settle four questions before a pilot's budget is released, not after results start to disappoint:
- Write the stop rule as a number, not a sentiment.
- Name an independent reviewer for the review seat - never the sponsor.
- Fix a hard calendar review date immune to extension requests.
- Name in advance where freed capital will go if the pilot is killed.
Missing any one of these, he argues, turns a written stop rule into a paragraph in a deck rather than something that actually happens on schedule.
Reframing the Resistance Boards Feel
Dr. Tebaa acknowledges that naming a reviewer outside the sponsor's chain often reads, to boards, as a judgment on the sponsor. He argues the opposite is closer to the truth: leaving the review seat with the sponsor does not protect that person, it puts them in an impossible spot, asking them to rule against a case they built, using a number they wrote, on a timeline they controlled. He notes that boards close to founding families in the Gulf and the Levant feel this pull particularly strongly, preferring review to stay inside relationships that already work - which, in his account, is exactly the setup that produced the expensive pilot rather than the cheap one.
Killing on Schedule as a Funding Mechanism
Dr. Tebaa's closing point reframes the entire exercise: killing a pilot on schedule, with its freed capital already named for a next destination, is not a defensive maneuver. In the composite case, the $254,000 freed when the cheaper pilot was stopped within two weeks of its missed review funded two other initiatives that shipped by that year's third quarter - capital that a slower kill would have kept locked inside a pilot already past its own target.
