Co-op Legal Services Scores Customer Calls With AI

Co-op Legal Services is using artificial intelligence to record customer phone calls about probate, wills and estates and give each call a percentage score, The Guardian reported on 8 October 2026.

The system is built around an OpenAI model. It is trained to check more than 50 separate features of each call, from wording to tone. Every call is scored. Managers receive pass and fail marks to review staff performance.

The monitoring began in the couple of months before 8 October 2026. It covers dozens of staff. The organisation uses the results in part to help staff improve sales.

The Co-op said the technology is solely a support tool, not a decision-maker. It said the tool helps managers support colleagues and improve the experience for clients. It described AI as a key aspect of its quality assurance, the process for checking advice is accurate and consistent, to help ensure clients receive empathetic expert guidance.

Recruitment material for its probate teams separately says those teams are supported by innovative AI tools and case management technology.

The broader context here is that one scoring system is doing two jobs at once. Quality checks in legal advice look for accuracy, consistency and empathy. Sales coaching looks for whether a call brings in business. Think of it as one report card grading both careful advice and selling skill. That gives managers a compact number, but it folds two different judgments into one.

Looking at what this means for professional services, the design of the checklist will carry weight. More than 50 measures means a detailed definition of a good call. Pass and fail marks give managers a filter for which calls to review. That can make oversight more consistent. It can also move judgment away from the adviser on the phone to the team that sets the measures.

In my view, practitioners will watch how the system is used day to day. How managers pair automated scores with human listening, what options advisers have to challenge a score, and what callers are told about recording and analysis will show whether the tool works as support rather than judge. For advice built on trust, those steps matter as much as the model.

For leaders weighing similar deployments, transparency and auditability will be central. A percentage is easy to share. The reasons behind it are harder to read without seeing the checklist, the weighting and the error rate. Managers need to know when to trust the label and when to listen themselves. Staff need the same clarity.