
Sauna pulls the exposure and loss triangles, drafts the reserving exhibit, builds the experience-study workbook, and writes the assumptions memo so your rate or reserve indication is documented before it reaches the committee.
What it does
Sauna queries paid and incurred losses from Snowflake, assembles the development triangles in Google Sheets with link ratios, and flags accident years where the pattern breaks from prior selections.
It compiles actual-versus-expected mortality or lapse data into a study workbook, shows the credibility weighting, and notes where exposure is too thin to support an assumption change.
Sauna writes the reserving or pricing assumptions memo in Google Docs, citing the data periods and judgment calls, ready for your peer review before it goes to the actuarial committee.
It manages the quarterly reserving deadlines and review sign-offs in Asana and drafts the status notes to the chief actuary so each exhibit clears on time.
Put Sauna to work on this.
Get started for freeIn context
Sauna shows up where you already work — the web app, Slack, email, iMessage, and Superhuman. It reads what it needs, does the task, and comes back with the draft for your approval.
Try it
The literal prompt for this job. Open it in Sauna and it picks up from there.
“Pull paid and incurred losses by accident year from Snowflake, build the development triangle with five-year volume-weighted link ratios, and flag any year that deviates from last quarter's selections.”
Plugs into the tools you already run — and thousands more, or any MCP server.
Good to know
No. It surfaces actual-versus-expected results and credibility, but every selection — link ratios, mortality, lapse — is yours. It documents which numbers came from data and which were judgment.
The losses come straight from your Snowflake tables with the query shown, and the link-ratio math is exposed cell by cell so you can tie every number back to source before you rely on it.
Sauna reads only what you connect, and acts only after you approve. Your workspace and its memory are yours, not training data.
Keep exploring