AI evidence categorisation: improving how evidence is managed in dispute resolution

Every day, we see around 8,000 to 10,000 pieces of evidence arriving from consumers and suppliers across our energy and communications businesses.

Bills, contracts, emails, meter photographs, call recordings and engineer reports can all form part of a case. Before our team can assess that evidence, they first need to understand what each file is.

At that scale, simply finding and organising the right evidence can create a significant amount of manual work, adding to the total time an investigation takes. 

Evidence Categorisation using AI is part of TAG’s wider focus on strengthening our capabilities and improving performance, using technology to help our teams work more efficiently and consistently.

Making evidence easier to navigate

To make evidence easier to find, each file is given a category that describes what it contains, such as a bill, call recording or engineer report. Previously, a colleague had to manually choose one category for each file from a relatively short list.

That didn’t always reflect the evidence itself. A recorded call about a bill, for example, can be both a recording and evidence relating to a bill. Categories could also be applied inconsistently, limiting how useful they were when searching through a case. 

The process has now been redesigned to create a more efficient and accurate system. A single file can carry several categories of evidence from a fixed list of around 30, reflecting the different types of evidence we receive. This categorisation is now applied by AI, helping our team search and filter evidence more effectively.

AI categorises. People decide.

The distinction is important.

These categories are sorting labels, not findings. They help identify what a piece of evidence is; they don’t determine whether it supports or weakens a case. 

Human context remains essential. AI may identify a document as a bank statement, for example, while a Dispute Resolution Executive may recognise that within the context of the case it is being used as evidence of payment. Categories can therefore be amended where necessary. 

Most importantly, AI does not make decisions about the case itself. That remains with our Dispute Resolution Executives.

Using technology where it adds value

Automatic evidence categorisation is designed to reduce the time spent manually sorting and searching through large volumes of evidence, improving consistency and lowering the risk of important information being overlooked. 

For Trust Alliance Group, this is driving the right environment to support our teams: creating the systems our teams need to work more effectively, while keeping human judgement at the heart of dispute resolution.

It also reflects the expertise of our Data and Lumin teams, turning that principle into practical improvements across our services.

Ultimately, it is about making complex processes simpler so our people can focus on what matters most: reaching fair and balanced decisions.