65% — the share of AI pilots in our sample that reached a production decision within twelve months. The remaining 35% weren't killed by a bad result; most were never formally closed at all.
21 of 26 — the number of stalled initiatives, out of the twenty-six we tracked that didn't reach a healthy first anniversary, that failed for the same root cause: no pre-agreed escalation path for low-confidence outputs.
3 to 1 — the ratio of enterprises that could produce an up-to-date architecture diagram to those that could name who was accountable for the decisions the diagram represented.
60% — the point in most legacy platform modernization programs at which our research found momentum most commonly stalls, typically as the easy, low-risk components run out and the load-bearing ones remain.
Full methodology and the underlying data behind each figure are available on request from the Institute's research team.