Insurance broker finds three AI use cases that pay back within a year

Situation

A commercial insurance broker in Frankfurt’s Westend, 35 staff and about €5 million in annual commission revenue, had been approached by three software vendors in six months. Each pitch promised a different AI tool. The two managing partners did not know which parts of their business would gain from AI, which would not, and what a sensible first step should cost. They wanted an independent view before signing anything.

The firm runs four departments: new business, policy administration, claims support and accounting. Most of the daily work is reading documents, comparing them, and writing letters and emails to insurers and clients.

Approach

We ran a three-week assessment. In week one we interviewed nine people across the four departments and sat with them while they worked, to see where the hours go. In week two we listed every recurring task that involves reading, comparing or writing, 27 in total, and scored each one on four points: hours per week, how repeatable the task is, whether the data it needs already sits in a usable system, and what could go wrong if the AI makes a mistake.

That scoring left 12 use cases that deserved a closer look. For each we estimated the hours saved per week, the likely cost of an off-the-shelf tool, and the payback period. Data protection was part of the score from the start: any use case that would send client health or financial data to a third-party model without a proper agreement was marked down.

In week three we wrote up the shortlist with a recommended order and presented it to the partners in a 90-minute session. The proposal also flagged that a firm of this size is likely eligible for the federal BAFA consulting subsidy, which the partners chose to look into with their tax adviser.

Result

In this scenario the partners received a scored shortlist of 12 use cases, a written recommendation to pilot three of them, and a recommended order for the first year.

The three recommended use cases were policy comparison summaries for renewals, first drafts of claims correspondence, and pre-filling of application forms from client documents. Together they address about 30 hours per week of staff time across five people, with an estimated payback of under 12 months each. Six use cases were parked for later, and three were rejected because the data protection risk outweighed the saving.

The partners approved the first pilot two weeks after handover. None of the three vendor tools they had been offered made the shortlist, which saved the firm an estimated €18,000 in first-year licences it would otherwise have signed for.

What it cost

The AI adoption strategy assessment cost €4,900, quoted as a fixed price before work started. That bought the interviews, the scored long list and shortlist, the payback estimates, and the partner presentation. Any AI subscriptions or licences the firm buys later are excluded. The work was covered by our 100% money-back guarantee: full refund on request within 14 days of delivery.

Every fixed-price engagement carries the 100% money-back guarantee. How the guarantee works →