The Stop Verdict: Why It Often Pays Most

Of the four verdicts, organizations resist Stop the most, and it often has the highest return. It has to be possible before the work starts.

By Harinderpal Hanspal on March 2026. Updated October 2026

An abandoned initiative costs its direct spend, then opportunity cost, damaged channel relationships and the credibility spent defending it. A defensible Stop before that commitment costs a small fraction.

A defensible Stop before the commitment costs a small fraction of an abandoned initiative A tall stack of four costs for an initiative abandoned after commitment: direct spend, opportunity cost, damaged channel relationships and spent credibility. Beside it a small block for a defensible Stop before commitment, with an arrow to the record that a Stop leaves behind. Credibility spent defending it Damaged channel relationships Opportunity cost Direct spend Defensible Stop Abandoned after commitment Stop before commitment A defensible Stop is a record, not a mood The hypothesis tested and the pass/fail criteria declared before the research Scored evidence from the buyers who would have had to say yes The dimensions on which that evidence failed A board can rely on it. The next hypothesis starts from it.
A defensible Stop before the commitment costs a small fraction of an abandoned initiative

The market is doing the same arithmetic. In S&P Global Market Intelligence's 2025 survey, 42% of companies were abandoning most of their AI initiatives, up from 17% a year earlier.

Teams rarely reach Stop alone. A team that spent twelve months building a hypothesis cannot credibly issue a Stop against it, and the sponsor who fought for the budget is badly placed to argue it should not be spent. That is an incentive problem, which is why the verdict needs someone with no stake in the answer.

A defensible Stop is a record, not a mood. It names the hypothesis tested and the pass/fail criteria declared before the research. It holds the scored evidence from the buyers who would have had to say yes, and the dimensions on which that evidence failed. A board can rely on it, and the next hypothesis starts from it.

Stop also has to be a real option from day one. If it is not, no verdict from the work can be trusted, Proceed included. In our reading, Gartner's forecast that more than 40% of agentic AI projects will be canceled by the end of 2027 amounts to a forecast of Stop verdicts issued years late, at post-commitment prices.

Go deeper: Too successful to stop, too unproven to scale covers what a defensible Stop contains, and what follows each verdict shows what the client keeps.