← All case studies

Case study / Insurance / Car renewals

85% less staffing required for renewals

An insurance agency realized it was spending too many resources on renewals. When it began working with Zuma, renewals required two full-time employees, plus hours of the CEO's day. Zuma built a custom renewals portal to tackle everything from new quotes to customer communications to policy issuance. Now, renewals are completely handled by the AI renewals agents, and the humans step in only when the Agents ask them to. The business results:

85%
Fewer staff hours required for renewals.Previously, two employees worked on renewals full-time. Now, one employee spends 30% of their working time on renewals, equivalent to 0.3 full-time roles.
Estimated90%
Less time from starting a renewal to completion.Based on feedback from employees and the CEO, rather than recorded before-and-after completion times. A shorter process means a better experience for customers.
One view
Of the renewal operation.The CEO once had to check five different systems of record to understand the status of renewals. Now, one centralized hub gives him a bird's-eye view of the renewal business.

The CEO was part of the renewal process

Car insurance renewals occupied two full-time employees. During active renewal weeks, the agency's CEO also spent at least a couple of hours a day finding out where things stood, fixing issues, chasing employees, and too often apologizing to customers when things fell between the cracks.

Time spent chasing renewals meant less time winning new customers, while existing customers received mediocre service. The CEO turned to Zuma to address both.

One portal, from renewal list to customer conversation

Zuma built a portal that automates the agency's renewal workflow, with employees reviewing customer communications and handling exceptions. It connects deterministic processes encoded in code with renewal agents that work 24/7 to give customers the best experience.

The renewal process
01

Prepare

  1. Identify customers due for renewal
  2. Gather customer and policy data
  3. Get quotes from insurance companies
  4. Compare quotes and select the best suitable option
02

Agree

  1. Present the quote and discuss with the customer
  2. Get customer approval
03

Complete

  1. Issue the new policy
  2. Charge the customer

We started with two goals:

  1. To fully automate the process
  2. To not only maintain but also improve the customer experience

To achieve this, we brought every renewal task and its context into the portal. Work done outside it leaves gaps in what agents can learn and automate.

This meant aggregating all the data needed to issue a renewal with a fantastic UX, which includes antiquated tools that don't even have an API to connect to. It also included centralizing all possible communication mediums with the customers to the portal itself.

Once everything was connected, we spun up three AI agents, each responsible for different things:

  1. Quotes agent
  2. Communications agents
  3. Improvement agents

The quote agent gathers data and finds the best quote under the agency's criteria. The communications agent handles emails, texts and calls. The improvement agent investigates corrections and turns them into lessons for the other agents.

When an employee corrects a draft, the improvement agent investigates why and extracts a lesson to prevent the same mistake from recurring.

Agents operate within rules enforced in deterministic code. For instance, the communications agent stays within the renewal's scope, and the quote agent works only with the agency's approved insurers.

The portal's north star metrics on the path to full autonomy

A portal that aims to fully automate what was previously a complex human flow has to be relentlessly data-oriented. Above all metrics, three North Stars guided our every action:

  1. End-to-end renewal completion time
  2. Percentage of flows that were completed without human rework (e.g., email drafts sent with no edits)
  3. Average human rework time per flow requiring rework

Different business lines have varying tolerance for error. With a digital insurance agency, we are balancing two forces. On the one hand, the first-principles thinking is that there should be zero tolerance for error. On the other hand, humans in charge of the flows already have a pretty high error rate.

The beauty of being data-oriented in a business that was never data-oriented before is that now this decision could be made on a dynamic, case-by-case basis by the CEO of the company.

Importantly, we see time and time again with our customers that, even as flows are becoming fully automatable, real successful businesses know that a customer must always be a click away from speaking to a person.

Protecting customer information

For an insurance business, private customer information should be protected above all else. Identifying details are redacted before data reaches the models, with a local model providing an additional check. The portal uses two layers to reduce the risk of exposing customer information:

  1. A deterministic redaction layer. Identifying details are removed or replaced with anonymous references before data is made accessible to the portal's agents. The mapping between those references and the customer stays inside the portal.
  2. A local model that reviews the abstract puzzle and flags if it thinks any information does leak.

When it comes to customer data, only the paranoid survive.

Renewals as an opportunity to grow customer value

Every renewal is a chance to understand a customer's needs and strengthen the relationship, and monetize more holistically. With customer information, communications and sales activity in one place, the agency can finally evaluate which renewal discounts help retain customers and where relevant additional coverage could increase the overall value of the relationship. Rather than judging each renewal in isolation, the team can consider its contribution to retention and profitability across the customer's policies.

You dream; we execute.

Book a free AI working session to map out how AI can transform your business and where to start.