Tide & Co.
Embedded copilots.
A copilot that lives inside a vertical SaaS — and earns its keep against a CSAT metric.
✺ — The problem
Tide & Co. sells software to independent insurance brokers. Their users spent half their day in spreadsheets reconciling carrier data. The company knew there was an AI play; what they didn't know was where to put it without making the product feel like a chat-bot tax.
Sector
Vertical SaaS
Year
2024
Duration
10 weeks
Team
1 Principal · 1 Designer · 1 Engineer
Stack
✺ — Approach
The same arc as every engagement — tuned to this problem.
Define · Where the keystrokes happen
We instrumented eight broker workflows for a week. The painful ones were not random — they clustered around two screens. That's where the copilot earns its keep, and nowhere else.
Build · Inline, not modal
The copilot lives in the cells, the headers, and the gutters. No chat panel, no '/' command palette, no over-eager suggestions. It surfaces only when the user is doing the painful thing, and goes silent otherwise.
Operate · CSAT as the production metric
Not 'tokens per query' or 'response latency.' We instrumented every copilot interaction with an unobtrusive thumbs-up/down and tied that to weekly CSAT. The model now optimizes against something real.
✺ — Outcome
Three numbers we’d defend in public.
−43%
time on reconciliation workflows
+18pp
user CSAT, six weeks post-launch
94%
of brokers used the copilot weekly within a month
“We were ready to spend a year on this. They told us in week one that 80% of the value sat in two screens. We shipped those, and the rest of the roadmap became optional.”
VP Product, Tide & Co.