Key takeaways
- Enterprises are past pilot fatigue; the 2026 question is operational integration, not capability proof.
- Risk and compliance functions have become the most active AI consumers — the opposite of the 2022 picture.
- Adoption velocity is now bottlenecked by data hygiene and access rights, not model availability.
The “are we doing AI?” board-level question has been replaced by a more useful one: “where is it actually working.” This piece looks at adoption patterns across financial services, retail, and healthcare.
Pilots are done {#section-1}
Most enterprises have a portfolio of 5–15 AI pilots that ran in 2024–2025. Through 2026, the surviving ones are being industrialised; the rest quietly retired.
Risk teams as buyers {#section-2}
The most striking shift: risk and compliance functions, once the most AI-skeptical, are now driving the largest internal procurement budgets. Auditable rule-extraction, transaction-monitoring augmentation, and KYC speed-ups are the killer applications.
What’s actually slow {#section-3}
Adoption velocity is set by data, not models. Companies with clean, accessible data are deploying in weeks; those without are quoted in quarters.