Organisations that would never sign an enterprise software contract without legal and architecture review are signing AI service agreements with far less scrutiny. That gap needs to close.
Most AI governance frameworks focus on what to prohibit. The ones that work focus on what to enable safely and that requires a different structure entirely.
Most organisations are running AI pilots. Far fewer have moved those pilots into production with any meaningful governance around them. The gap between experimentation and governed adoption is where AI value gets lost and where AI risk gets created.
One anonymised financial-services case example cut its release cycle from 8 months to 5 weeks. Not their development cycle. Their release cycle. From the moment someone says “we want to build this feature” to the moment it’s live in production. Eight months. That’s longer than a human pregnancy. But their code only took 3 months…
