Most AI conversations in insurance skip past the hard part - how the work actually gets done. I focus on the messy middle: document handling, submission intake, renewal prep, and the dozens of places where smart people lose time to manual processes.
Insurance doesn't need more thought leadership about AI. It needs people who've actually done the work - sat across from a client, built a program, managed a book, run an operation - and can see where these tools genuinely help.
I've worked across production, advisory, executive leadership, and entrepreneurial roles. I founded and ran my own agency, sold it, and went on to leadership roles at a top global brokerage platform and in PE-backed environments. Along the way, I kept noticing the same thing: too much of the real work is buried in PDFs, emails, loss runs, and spreadsheets that people reassemble by hand every time.
Once language models got good enough to be useful - not perfect, but useful - it felt obvious that there was real opportunity to fix some of that. Not to replace people, but to get them to the right information faster.
Three things determine whether AI in insurance actually works or just looks good in a demo.
I start from the actual work. What decision is someone trying to make? What information is missing? Where does the process break down? Then I figure out where AI can help - intake, research, servicing, renewal prep, document review, knowledge retrieval.
Most AI conversations overemphasize what the model can do and underemphasize what it takes to make it usable: process design, review steps, role clarity, escalation paths, who owns the output, and how you'd measure whether it's actually working.
Insurance is regulated for good reason. If you're deploying AI in this space, you need real controls - review expectations, audit trails, privacy boundaries, risk tiers, and clear accountability. Anything less and you're just hoping nothing goes wrong.
Each project starts from a real insurance task. I'd rather build a rough working tool and learn from it than spend time perfecting a strategy deck.
A tool for pulling useful information out of insurance documents - policy excerpts, submissions, loss runs - and answering structured questions with sources you can actually check.
View projectGenerates account briefs, renewal prep notes, and meeting materials from client inputs. The kind of work that eats hours every week and usually involves reassembling the same scattered information.
View projectA working framework for sorting AI use cases by risk level, defining who reviews what, and building the controls you'd actually need before deploying anything in a regulated environment.
View projectI write about where AI can create real value in insurance and where the claims outrun the reality. Less thought leadership, more operator perspective.
The AI conversation has been carrier-centric for years. But a huge share of the friction lives upstream, inside brokerage operations.
Read essayMost firms have a policy. Very few have a governance model that actually tells you what to do when you're deploying something.
Read essayI'm building a public portfolio of applied AI work - tools, frameworks, and writing - focused on the insurance workflows I know best. Everything lives on GitHub. The goal is to show what I can build, not just what I can talk about.