Insurance
The AI Office for agencies and MGAs.
Underwriters spend roughly 70% of their time on work that is not underwriting. About 40% of it is purely administrative.
Meanwhile commission revenue per account is falling while service work per account is not — CIAB put Q2 2026 premiums down 2% overall and down 6.3% in commercial property.
If any of this sounds familiar
We would start by looking at Applied Epic or AMS360 — though the real bottleneck is IVANS download and emailed PDFs.
- Loss runs arriving from every carrier in a different format on a different cadence
- Commission statements reconciled line by line, by hand
- Submission packets assembled from scratch for each market
- A mandatory human exception path that never gets smaller
Being straight with you
We have not done this in your industry yet
We have done the identical problem in a different one, and we would rather say that plainly than dress it up.
This is barely an analogy, and in one place it is not one at all. N carriers, files in different formats, no shared schema, a mandatory human exception path, a blocked downstream process — in benefits administration it blocked billing; for you it blocks underwriting, and partly the same institutions are sending the files. We also built claim validation and duplicate-payment detection on that same client’s insurance operations, which is insurance work rather than a parallel to it.
Where we would start
One workflow. Four to six weeks. Fixed fee.
Loss-run and commission-statement normalization, running on your shared Microsoft 365 mailbox — no agency-management-system credential needed on day one.
Before we start, we agree in writing the number the build has to hit. If it does not hit it, we keep working at no additional cost until it does.
$6,000
credited in full against month one of any retainer started within 30 days
Or go straight to a retainer from $2,500/month. Month to month, 30 days’ notice.
The build order
Ranked by what a wrong output costs
We start where mistakes are cheap and visible, and earn the right to touch money. If you ask us to begin at the bottom of this ladder — and people often do — we will explain the sequencing rather than agree to it.
A wrong output costs a re-read. Start here, always.
Loss-run normalization to one schema · submission packet assembly · shared-mailbox extraction, classification and filing · carrier appetite pre-screen
A wrong output creates re-work and a client conversation.
COI issuance behind a hard approval gate · policy checking against binder and application · commission statement variance · duplicate payment detection before release, which we have built and run
Last. We would not scope these in a first engagement, and we would not put them in an email.
Coverage determination or gap analysis, which is advice · premium calculation and rating · claims payment authorization, reserve setting and denial
The question you are about to ask
“Which agency have you done this for?”
No agency, broker or MGA — and we would rather say that plainly than let you find out later. What we have done is fourteen engagements for a benefits administrator on the identical problem: many carriers, incompatible formats, a human exception queue that never shrinks. That included claim validation and duplicate-payment detection on their insurance operations. So the domain is not new to us. Your channel is. If that distinction disqualifies us, it is a reasonable call to make.
Your data
Your data never leaves your tenant.
We do not hold your data, because we do not build on our infrastructure. We build inside yours.
- Built inside your own Microsoft tenant
- Entra ID role-based access through security groups
- No static or local accounts
- MFA enforced
- SFTP with IP whitelisting for partner feeds
- Sensitive data masked in every non-production environment
Continuity
One client has renewed us fourteen times since 2020.
A benefits administrator has brought us back for fourteen sequential engagements since 2020 — eligibility reconciliation, invoice automation three separate times as the requirement changed, pre-certification workflows, a data-quality programme — and we still operate their platform today, with alerting, weekly. Every one of those fourteen ended with them asking for the next thing. That is the part worth weighing: fourteen engagements is not one relationship, it is thirteen separate decisions to keep us, each made by someone who could have stopped. Six years is the honest answer to the question behind every retainer decision — will you still be here in year three.
Twenty minutes, and you will know.
We ask what your team spends its week on. If there is nothing here worth doing, we will tell you that.