I'm an insurance professional who got distracted by AI and never looked back.
25+ years in insurance. Two years seriously building with AI. The last two taught me more than the previous 24 combined.
On one hand: underwriting, risk, and the business of insurance — two decades of it across markets, products, and cycles. On the other: building automation systems and intelligent workflows that let a small team do what used to need five times the people.
The thing that changed everything wasn't the technology. It was having something that could teach me — from the mundane to the most complex — without judgment, without gatekeeping, without waiting for someone else's availability. Zero dependency on others. Zero "let me check with the developer." Zero "we'll need to bring in a consultant."
Ask it anything. It explains it like you're an idiot. You build it anyway. It works.
I've seen this across both ends of the spectrum. Google Workspace works for small businesses. Microsoft Copilot Studio works for enterprises. Two stacks, two different segments, one pattern: making intelligent automation is now possible for people with real problems and no patience.
The Google builds run on Sheets, Apps Script, and the Claude API. The enterprise builds run on Copilot Studio, Power Automate, and Azure. Both attempted — and increasingly succeeded at — with AI as the co-builder.
No dev team, no six-figure budget. All that's needed is understanding how businesses actually work, a powerful platform, and an AI that explains things like I'm five.
Turns out that's enough.
Every piece starts with something that's bothering me.
A number that doesn't add up. A narrative that feels too clean. A question nobody in the room is asking. Something I've seen in 25 years of insurance that contradicts what the textbook says. A build that worked in a way I didn't expect.
I bring that to Claude. We pull the primary material — data, filings, research, records, whatever the question needs. We build the thinking together. I challenge the outputs. Claude challenges my assumptions. We go back and forth until the answer is honest or the question gets harder.
The result is what you read here.
It spans everything I'm curious about — economics, insurance, AI, startups, leadership, whatever I'm turning over this week. The subject changes. The method doesn't.
No political affiliation. No institutional position to protect. No advertiser to please. No agenda except the question I started with.
Claude has no vote, no party, no editor, no quota, no client. That's why this works.
Disclosure
This site uses Claude (Anthropic) as a research and writing collaborator across all subjects. Primary sources are cited where used. All editorial positions, opinions, and conclusions are those of HR Srinivas. Claude's outputs are reviewed, challenged, and edited before publication.
This is AI-assisted thinking made public. Not AI-generated content.