Kay AI

Kay AI takes the opposite approach to most 'AI for insurance' tools. Rather than bolting a chatbot onto an AMS or selling a workflow builder your ops team has to configure, it asks you to hand over the SOPs you already have — documents, spreadsheets, email chains, and the procedures that live in people's heads — and turns them into a living procedure document in days. Then it executes: navigating carrier and broker portals, filling forms, pulling carrier downloads, running policy checks, processing endorsements, and writing AMS activity. On the homepage Kay names six workflow families it covers: certificates, renewals, policy checks, carrier downloads, AMS activities, and endorsements. The human-in-the-loop design is the part that separates it from generic RPA. When Kay hits something it isn't sure about, it pauses and asks on Microsoft Teams or email, remembers the correction, and improves the next run — no retraining project, no model work from your side. The vendor also advertises a real-time audit trail of every run, which is the feature that matters most when an E&O review lands on your desk. Scale and reliability claims are specific rather than vague: 24/7 operation, hundreds of tasks in parallel, and 'no queue, no delays.' Customer quotes on the site give you numbers to test in a pilot rather than adjectives — 93% time saved on renewal reshops, certificate issuance learned in two weeks against a stated six-month staff training baseline, and a deployment live in four days. There's also a '2x quality' quote from a VP of Agency Operations, though the homepage doesn't define how that was measured, so treat it as a customer testimonial rather than a benchmark. Where Kay is deliberately narrow: it is insurance-only, so nothing here transfers to another vertical. It checks in through Microsoft Teams and email rather than Slack, which is a real friction point for shops standardized on Slack. It is not fully unattended — by design it stops and asks when uncertain, so teams expecting straight-through processing with zero human touch will be disappointed. And the quality of what Kay learns depends on how well your processes are documented at the start; if your procedures only exist in one person's head, the ingestion step is where the work goes. Integration coverage is also narrow on the public pages: Applied Epic and Sagitta for AMS, Microsoft Teams for the human loop. The company is based in Brooklyn and states it was built in NYC, with a founder note arguing that insurance has been stuck choosing between people who don't scale and software that doesn't do the work.