Understanding the challenge
The MGA knew AI was coming for underwriting, and it had the budget to move, but no one could say where to begin. Competitors talked about AI underwriting, vendors pitched a dozen different tools, and the firm's leaders worried about the things a regulated insurer has to worry about: model risk, explainability, and legacy systems that could not simply be ripped out. "We did not want to buy a tool and hope, we wanted to know where AI would actually pay off and where it was too risky," the firm's COO said. So the MGA stayed on the sidelines, spending nothing and gaining nothing, while its underwriters kept doing by hand what software could have helped with.
Our approach
The MGA did not need a build. It needed clarity on where AI would create the most value before spending a dollar on tools. We ran our AI Strategy engagement on the principle the firm shared: value before hype, and strategy before software. We started by understanding the MGA's workflows, systems, team, and goals, then identified where AI could reduce manual work, speed decisions, or sharpen risk selection:
• Use-case discovery across submissions, underwriting, pricing, and claims, surfacing where the real manual load and the real opportunity sat.
• Opportunity prioritization that ranked every use case by value and effort, so the firm could see what to do first and what to leave alone.
• Tool recommendations matched to the MGA's actual systems and budget, not a generic vendor list.
• A governance and risk framework built for a regulated insurer, so model risk, explainability, and oversight were designed in from the start.
• A prioritized roadmap with implementation steps and clear next actions the team could execute.
Technical innovation
The core of the work was a prioritization model that scored each opportunity on value and feasibility together, weighing the size of the manual load against data readiness, integration effort, and the model-risk and regulatory exposure of automating it. Because we run more than fifty operational AI tools inside our own agency, the recommendations came from what actually works in production rather than vendor marketing, and every high-value use case was pressure-tested against the insurer's explainability and oversight requirements before it made the roadmap. The result was not a wish list, it was a sequenced plan the MGA could hand to a vendor or an internal team and start executing.
Outcome
In about six weeks, a year of hesitation became a board-approved plan. The roadmap ranked the firm's AI opportunities by value and effort, named the quick wins it could start immediately, and drew a clear line around the ideas that were not worth the regulatory risk yet, all inside a governance framework the underwriters and compliance team were comfortable with. For the first time the MGA knew exactly where to invest and where not to. "We finally have a plan we trust instead of a pile of vendor pitches," the COO said.
