Outreach depended on who was making the calls
In a multi-line agency, almost every customer could use another policy. The hard part is knowing who to talk to, about what, and when. At this agency, only the CEO really knew.
The CEO could tell when a customer was ready to hear from the agency and when a call would land badly. But one person can't sit next to every agent or check every decision about who to call next.
So each agent made their own calls. Some reached the right customer at the right time. Plenty didn't. Nobody was doing a bad job. The agency just had no way to turn one person's judgment into a rule everyone followed.
The hub
This didn't start as a revenue project. The goal was efficiency, automating the work that filled the team's day. The revenue came later, from doing that work the way the CEO would have.
We built the hub on top of the CRM the agency already used, so nobody had to move to a new system. It took over three jobs that used to depend on each agent. It drafts every outreach message, so agents start from a draft instead of a blank page. It chooses which customer comes next. And it decides when to call.
Underneath all three is the order of operations. We encoded the CEO's judgment as the full sequence: who comes first, who can wait, and when to reach out. The system runs on it, so nobody has to remember it from a training session.
Now the CEO's judgment shapes every agent's next call, whether or not the CEO is in the room.
The car-insurance discount was the way in
Agents need a reason to call. Here, it was the car-insurance discount.
Telling someone they can save money is an easy call to make, and customers are happy to take it. Once that conversation is going, the agent can bring up life insurance.
Not every discount call should turn into a life insurance conversation, and timing matters. The hub decides who gets called and when, using the CEO's priorities.
Data and people
The AI was the easy part. Two other things took most of the work.
The first was data. The old and new systems disagreed, and a system that picks who to call is only as good as its records. A wrong record means calling the wrong person about the wrong policy. We wrote sync logic that treats the insurance companies as the source of truth. When the systems conflict, the insurer's record wins.
The second was people. Agents were used to deciding for themselves who to call. A system that picks for them only works if they follow it. Getting the team to let the hub drive took as much effort as writing the code.
What comes next
With the CEO's judgment in a system, the agency can see what works and change it. If a priority turns out to be wrong, they change it once and every agent's queue changes with it. The car-insurance discount was the first way in. Other lines of business could use the same setup.