When AI initiatives underperform, the leadership response usually targets two things: "the team isn't adopting" or "the tool isn't powerful enough." Both diagnoses are almost always wrong. The actual cause sits upstream — usually in decisions made before the rollout even started.
"The team isn't adopting." Translation: "we deployed something, expected enthusiasm, got resistance." This usually leads to more mandates, more training, more pressure — none of which fixes the underlying problem.
"The tool isn't powerful enough." Leads to vendor evaluations, switching costs, and 6 more months of "we're evaluating alternatives." The new tool will produce the same outcome because the problem wasn't the tool.
Both diagnoses are visible (you can point to them in a board meeting). The actual cause is invisible — which is why it persists.
The real cause is upstream: the rollout was scoped against the wrong success metric, or against no success metric at all.
When you say "we're going to roll out AI to the marketing team," what does success look like? If you can't answer in specific measurable terms, the rollout is doomed before it starts — not because of execution, but because there's nothing to align on.
The team can't tell what good adoption looks like. The leadership can't tell whether it's working. The vendor can't demonstrate value. The whole effort drifts toward "well, we tried" with no clarity on whether it worked.
Scoped wrong: "Deploy AI to the marketing team to improve productivity." → No measurement, no accountability, no clarity.
Scoped right: "Within 90 days, deploy 3 AI workflows that take marketing-team blog post production time from 8 hours per post to 2 hours per post. Measure weekly. Specific person owns the outcome." → Clear, measurable, accountable.
Scoped wrong: "Roll out Claude to the sales org for productivity gains." → Sounds reasonable, achieves nothing.
Scoped right: "Within 90 days, AEs spend at most 10 minutes per account on research (down from current 90 minutes), maintain or improve outbound reply rates, and ship 50% more sequences per rep." → Clear baseline, clear target, clear measurement.
Before any AI rollout starts, force clarity on three things:
1. The baseline. What's the current state of the metric we expect to move? If we don't know, we can't tell whether it improved.
2. The target. Specific, measurable, time-bounded. "50% reduction in proposal turnaround time within 90 days," not "improved efficiency."
3. The accountable owner. One named person responsible for hitting the target, with authority to make decisions about the rollout. Distributed accountability = no accountability.
Without these three, no amount of training, vendor switching, or "AI culture" work will produce results. With these three, even mediocre execution typically produces meaningful gains.
For the operating-level version of this discipline, see the 90-day AI rollout playbook. For why most rollouts fail (broader diagnosis), see why most AI implementations fail.
— Bill Colbert, Treetop Growth Strategy