The adoption-to-value gap: why AI spend isn't landing

Two numbers, widely reported across 2025 industry surveys, sit awkwardly next to each other: a large majority of organisations say they have adopted AI tools, and a small minority say they can point to realised value. Treat the exact percentages as indicative — every survey measures something different — but the shape is consistent and worth taking seriously.

Why the gap exists

Value from AI in delivery is bottlenecked by the slowest governed step, not the fastest ungoverned one. If a change still needs three approvals, a security review, and an audit trail before it ships, then generating it in ten minutes instead of two hours moves nothing that leadership can see on a P&L.

• Measurement. Most "productivity" is measured at the keystroke, where AI helps most, not at lead-time-to-value, where the constraint actually is.
• Trust. In regulated settings, an output you cannot evidence is an output you cannot use — so it queues behind human verification anyway.
• Governance debt. Faster generation with unchanged decision and evidence capacity simply grows the backlog of things waiting to be governed.

Closing it

Our read: the gap closes when you move the governance work into the flow of delivery rather than bolting it on afterwards — decisions captured as they are made, evidence produced as a by-product, assurance running at runtime. That is a deliberately unglamorous answer, and it is the one that survives contact with an auditor.

Evidence label: our analysis. The underlying adoption/value figures are drawn from public 2025 industry surveys and should be read as indicative of direction, not precise.