AI Is Great at Exploring, Not So Great on Rails — Mark Hay

Mark Hay, CTO of TextQL, in a 22-second excerpt from a TwoSetAI conversation about AI and business data.

The tension

• Teams build heavily engineered controls because a missed metric definition or a query against the wrong data source can cause real harm.
• Hay says AI is particularly good at figuring things out and exploring; it is less effective when forced onto rigid rails.
• The clip raises a design trade-off, not a tested prescription: constrain high-risk outputs without eliminating the exploratory work that makes an agent useful.

“AI is really, really good when it’s trying to figure stuff out and explore things, and it’s not really that good when you’re putting it on rails.”