Acolite Is Building AI Coworkers for Insurance Agencies That Are Running Out of Actual Coworkers

The Macro: Insurance Has a People Problem

The insurance industry is facing a workforce crisis that makes tech’s talent wars look quaint. The numbers are stark. Seventy-three percent of account managers are over 60 and heading toward retirement. Replacing them takes four to five months per hire, assuming you can find someone willing to do the job. Younger workers are not exactly lining up to spend their days processing certificates of insurance and mapping statements of values into agency management systems.

This is not a new problem. The industry has been aging out for a decade. But the pace is accelerating, and the operational complexity is not getting simpler. An insurance agency runs on paperwork, emails, and data entry. A single submission to a carrier involves pulling client data from the AMS, formatting it correctly, attaching the right documents, and sending it to the right underwriter at the right carrier. Issuing a certificate of insurance means pulling policy details, filling in the right fields, and making sure the certificate holder information is accurate. These are tasks that require attention and familiarity with the systems but not deep expertise.

The incumbents in insurance technology are mostly focused on the carrier side or the distribution side. Vertafore and Applied Systems own the AMS market. Bold Penguin and Tarmika handle quoting. Duck Creek and Guidewire build core systems for carriers. But the day-to-day operational grind inside an independent agency, the email processing, the certificate generation, the data mapping, has been largely untouched by automation.

A few companies have tried. Indio Technologies (acquired by Applied Systems) digitized applications. Agency Zoom automated CRM workflows. But nobody has built what I would call a true operational copilot for the agency back office. The tasks are too varied, too context-dependent, and too tied to specific AMS configurations for simple workflow automation to handle them.

That is the setup for an AI-native solution, and the timing lines up. Language models are now good enough to read unstructured emails, extract structured data, and take actions in external systems. The question is whether someone can build that into a product that works with the messy reality of insurance agency operations.

The Micro: Email In, Work Done

Acolite is a three-person team based in New York, part of YC’s Spring 2025 batch. The company was founded by Daniel Siryakov and Barak Ben Noon.

The founding team is well-matched to the problem. Daniel is a former ML engineer and AI product manager. Barak previously led AI feature development at a major tech company. They have the technical chops to build the AI side of this. The question, as always with vertical AI products, is whether they have the domain depth to navigate insurance-specific workflows.

The product pitch is tight. Forward an email to Acolite, and it handles the rest. That “the rest” includes processing submissions, issuing certificates of insurance, and mapping statements of values across any agency management system. The “across any AMS” part is ambitious. Vertafore’s AMS360, Applied’s Epic, HawkSoft, EZLynx. Each has its own data model, its own API (if it has one at all), and its own quirks. Building integrations that work reliably across all of them is a significant engineering challenge.

The email-forward interface is smart. It meets agents where they already work. Insurance agency operations run on email. Submissions come in via email. Certificate requests come in via email. Building an AI teammate that lives in the email flow rather than requiring agents to learn a new tool reduces adoption friction considerably.

I am curious about the accuracy requirements here. Insurance is a regulated industry. A certificate of insurance with the wrong policy number or the wrong additional insured is not just an inconvenience. It is a liability issue. The margin for error on these documents is essentially zero. That means Acolite needs to be right nearly every time, or it needs a human review step that is fast enough to not negate the time savings.

The company was founded in 2024, which gives them a head start on training data and integration work before the YC batch. At three people, they are lean, but the focus is narrow enough that three people can build a real product.

The Verdict

I think the market fit here is obvious. Insurance agencies are drowning in operational work, losing the people who know how to do it, and struggling to hire replacements. An AI system that can handle certificate issuance and submission processing accurately is worth real money to these agencies. The willingness to pay should be high because the alternative is hiring a human at $50,000 to $70,000 per year.

The risk is accuracy at scale. Every AMS is different. Every agency has its own workflows and preferences. The long tail of edge cases in insurance operations is enormous. A system that works perfectly for standard commercial auto certificates might break on complex excess liability certificates with multiple additional insureds and specific waiver of subrogation language. Getting from 90% accuracy to 99% accuracy in this domain is likely to be the hardest part of the business.

At 30 days, I would want to know how many agencies are actively using it and what their error rate looks like. At 60 days, the question is whether agencies are expanding usage beyond the initial workflow or keeping it boxed to one task. At 90 days, I would want to see retention. Are agencies sticking with it, or are they trying it and going back to manual processes because the edge cases are too costly?

The demographic tailwind is strong. The product concept is sound. If the execution holds up under the weight of real-world insurance complexity, this is a business that could scale across thousands of independent agencies.