Case studies.

The problem

Service requests arrived by email across many agencies. Someone had to read every message before work could start, and about two thirds of what came in needed no action: carrier notices, duplicates and spam.

What we did

1. Broke service work into defined tasks. A certificate request, for example, runs as six steps.
2. AI reads each incoming email in under 10 seconds, sorts it and routes the real requests by license and skill.
3. Routine tasks that follow a known path are done by AI. People take the exceptions and anything that needs a license.

The result

About 44% of completed service tasks are now done by AI, steady at 42–47% each month from March to June 2026. Around 40,000 tasks a week are sorted and routed with no one touching them first.

The problem

Nobody could say what one service task cost. Time disappeared between systems, and tasks got stuck with no record of why.

What we did

1. Timed the active work on every task, so slow steps showed up by task type.
2. Standardised the steps for each task type and moved the routine share to AI.
3. Tracked every stuck task by cause and fixed the most common causes first.

The result

Cost per task fell by roughly half between the March 2026 launch and June. People now spend a median 3.8 minutes on a task, down 16%. Stuck tasks fell from 435 to 348 a day and kept falling each week.

The problem

Each agency's book was spread across an AMS, carrier downloads and older systems. The same customer showed up several times, and systems disagreed on the same policy.

What we did

1. Pulled the books of 59 agencies from 9 systems into one place.
2. Matched records that belonged to the same customer and the same policy.
3. Set rules for which value wins when two systems disagree, and flagged what was missing.

The result

780,000 raw customer records came down to 333,000 real customers. About half of all customers existed in more than one system. On the same policy in two systems, the carrier didn't match 32% of the time. About 1 in 7 customers had no email and no phone on file.

The problem

CSRs moved small pieces of data between carrier portals and the AMS thousands of times a day: a policy number, a name, a coverage detail. Each one was slow, and a mistyped digit meant rework and a callback.

What we did

1. Recorded how work moved across portals and the AMS, using field labels only, never client data.
2. Built a one-click fill for the service form from data already in the system.
3. Sped up the full copy and mapped the busiest carrier portals first, choosing the next ones from real use.

The result

A full client file now copies in about 7 seconds instead of over a minute, roughly 9 times faster. Re-typing data the system already had stopped. 32 carrier portals are mapped, and the list grows from what the team actually uses.