Strategy and Enterprise with AI
AI is replaying the SaaS cycle in reverse. “ChatGPT for everything” is the easy part.
A mid-market insurer rolled out a generic AI chat tool to help adjusters write claim notes and customer emails. Month one looked like a win - about 15 minutes saved per claim.
By month three, cycle time was up 2 days. Every adjuster had a different prompt, different wording, and the same loss got explained five ways. QA kicked files back, legal wanted tighter language, and managers started policing prompts instead of leakage.
Durable AI in claims will not feel like a chat box. It will feel like one step inside the claims system: take photos on mobile, pull facts into the right fields, suggest reserves, then draft the note in your house style with the evidence attached. No one thinks about models. They think about outcomes.
This week the industry is fixating on audio-first AI interfaces. That only accelerates the point: the winners are the teams that understand one boring workflow deeply enough to make the AI disappear.
It makes me wonder whether most “AI strategy” is really a workflow ownership problem.