An anonymous real example
A participant described routinely using AI at home to help challenge companies when goods failed to arrive or arrived broken. Drafting a formal complaint and working out what questions to ask had become an ordinary use of the tool. The Building Trust insurance story takes this further: residents organised evidence and challenged explanations in a prolonged dispute, using AI alongside professional advice and collective effort. The customer on the other side of the process was better prepared to keep going.
The assumption worth revisiting.
A US study of more than 1.13 million financial complaints estimated a 6.9 percentage-point increase in favourable-relief probability associated with LLM adoption. That is evidence about complaints in a specific setting, not a universal prediction for your business. The wider question is strategic: if customers could understand their options, assemble their evidence and act with much less effort, would your offer still earn the same return? Complaint workload, redress, renewal and switching could change in different ways. Some exposure may sit in service quality; some may sit inside the business model. Use your own economics rather than applying a headline percentage to profit.
Are customers choosing to stay, or finding it too hard to act?
Give the capacity a purpose.
Bring Customer, Finance and Revenue together to find where inaction supports earnings, resolve avoidable friction and build a stronger reason for customers to stay.
How this fits the capacity allocation →
Use AI to challenge your first interpretation
Explore this question without assuming our first interpretation is right: Are customers choosing to stay, or finding it too hard to act? Offer competing explanations, identify assumptions and suggest evidence we could obtain from the people doing the work.
Treat the answers as possibilities to test with people and evidence.
Go deeper into the evidence.
Nature Human Behaviour · Research study
An analysis of 1,134,512 US financial complaints estimated a 6.9 percentage-point increase in favourable-relief probability associated with LLM adoption (95% confidence interval: 4.9–8.9 points).
An instrumental-variable estimate, not a UK insurance trial. Favourable relief includes outcomes beyond cash payments; do not turn this into a universal P&L percentage.
UK Government · Government estimate
The government estimates UK consumers spend £1.6bn a year on unwanted subscriptions.
A market exposure estimate, not a measured effect of AI or a company-level revenue forecast. This resource is used for the estimate, not as legal guidance on implementation dates.
video · 4 min
AI lawyers are here – can you trust artificial intelligence for legal help?
University of Sheffield
A short introduction to the opportunity and limits of AI-supported legal help, connecting customer empowerment with responsibility for accurate decisions.
Listen or watch for: What can a better-prepared customer now question, and where will they still need expert support?
About this resource
University and legal-support practitioner perspectives, including student tests; not evidence of a measured increase in successful legal claims.