Why doesn't a large deployment number settle the debate on AI readiness?

Azati points out that any well-funded company can generate hundreds of use cases with enough investment. The real test is whether use case #260 is as safe, governed, and auditable as #1, which depends on unglamorous data work rather than the count itself.

What pattern has Azati's CTO seen repeat across previous technology waves?

Azati's CTO has watched new technology layers arrive promising to change everything, only to see success hinge on unglamorous foundation work. That foundation work usually didn't get done, which is why the pattern kept repeating across waves.

What does responsible AI leadership look like according to Azati?

Azati notes that keeping focus on governance and human judgment in a regulated environment signals someone who has shipped before. This approach contrasts with leaders who sell a revolution without addressing the underlying data risk.

What is the actual prerequisite for scaling AI use cases safely?

Azati identifies AI-ready data engineering and governed integration as the real prerequisite for scaling AI use cases. Without this foundation, organizations end up deploying agents on data they already admit they don't trust.

Do 260 AI Use Cases Mean Anything If You Don't Trust Your Data?

Back to quick insights

August 19, 2026

Do 260 AI Use Cases Mean Anything If You Don't Trust Your Data?

Andrew Babkin

Managing Director & CTO, Azati

Quick answer: Prudential says 260+ AI use cases and 2,300+ people on agentic tooling, with Bob Bastian now running it as Chief Data and AI Officer. Good for them, those are real numbers, not press-release fantasy, and we’ve all read enough of those to tell the difference. But here’s the number I actually sat with this week: a Teradata-commissioned survey of 1,000 leaders found 77% say 20% or less of their data is genuinely ready for agents to use. Everyone is racing to deploy agents on top of data they already admit they don’t trust. So no, 260 use cases isn’t the flex.

One view: use-case count is momentum

Lots of live use cases and thousands of people using agentic tools is a real signal an organization has moved past PowerPoint.

The other view: it’s vanity until the data holds

Any well-funded company can generate 260 use cases. The mind-blowing part is when use case #260 is as safe, governed, and auditable as #1 – and that depends on the boring data work underneath, not the count on the slide.

Our read

I’ve watched this movie before. Every few years a new layer arrives that’s going to change everything, and every time the thing that decides success isn’t the layer, it’s whether the unglamorous foundation work ever got done. It usually didn’t. Bastian, to his credit, keeps the focus on governance and human judgment in a regulated environment – that’s someone who’s shipped before, not selling a revolution. The prerequisite is AI-ready data engineering and governed integration.

FAQ

Does a high AI use-case count mean a company is ahead?

Not by itself. Use cases are easy to generate – the hard part is making every one governed and auditable, which depends on data readiness.

Why isn’t enterprise data ready for AI agents?

Most of it lacks the context, lineage, and governance agents need – 77% of leaders say 20% or less of their data qualifies. Fixing that is data-foundation work, not a model choice.

Racing to deploy agents on shaky data?

Fix the foundation first

Adapted from a LinkedIn post by Andrew Babkin, Managing Director & CTO at Azati – read the original. Sources: Prudential Financial newsroom; Teradata / Wakefield Research, 2026.