4 Working AI Tools in a Few Weeks to Cut Manual Customer Support Work for a Major UK Water Company
A short proof of concept programme with a major UK water and waste company, where we built AI tools that check customer evidence before it is submitted, read claim forms for specialists, and score every support call against the company's own quality standards.
Company Overview
The client is one of the larger water and waste companies in the UK, serving millions of households and businesses. Its customer operations teams handle a high volume of applications, claims and calls every day, many of them from customers who need financial or practical support and who should get a quick, accurate answer.
A lot of that work was still manual, so specialists spent time chasing missing evidence, re-keying data from forms and checking details across systems before a case could move, which is exactly the kind of work AI and simple automation can take off their plate.
Problem Statement
Customers often submitted applications with incomplete or incorrect evidence, so specialists had to chase them and the case went round the loop again. Claims arrived both online and on paper and had to be read and keyed in by hand, with manual look ups against other systems before anything could be calculated. Call quality reviews could only cover a small sample of calls, which meant coaching was based on a fraction of what advisers actually did on the phone.
Outcome Delivered
In a few weeks the company went from a list of use cases to working proofs of concept for evidence checking, claim intake and call quality scoring, a process fix already live with the team, and a clear, step by step plan for each tool, starting with quick wins that need no AI at all.
Executive Overview
We ran a short, sprint based proof of concept programme with the customer support teams of a major UK water and waste company, focused on the work that slows specialists down: checking the evidence customers send in, reading claim forms by hand, and reviewing calls for quality. Each use case was built as a working tool, inside the client's environment where that was possible, so the business could see it running on real processes rather than on slides.
The programme delivered working proofs of concept for guided evidence upload, automatic application validation, claim intake from online and scanned paper forms, and call quality scoring with coaching for team managers. Two further ideas, an early warning operations view and real time support for advisers on live calls, were taken to a working demo and parked with a clear list of what it would take to move them forward.
Our Approach
We worked in short sprints, each one taking a use case from requirements and a journey map through to a working proof of concept and a handover specification for the client's own technical teams. Where it made sense we built inside the client's environment so it could be inspected, and where customer data was involved we used sample data until data protection approval was in place.
For every use case we looked at the whole process before reaching for AI, because adding a model on top of a broken step does not release the benefit. That is why some of the recommendations are simple process and website changes, done first, with the AI tools switched on once those changes have shown results.
Challenges
Challenges Faced
Every engagement starts with a clear-eyed view of what was actually getting in the way, because you cannot fix what you cannot see. Here we set out the real constraints the client was working within, the costs and risks they were carrying, and the gaps that needed closing before any meaningful progress could be made.
Let's Transform Your Enterprise Today
Send an EnquiryEvidence that is wrong first time
Applications often came in with missing or incorrect evidence, creating rework for specialists and delays for customers who needed help.
Forms keyed in by hand
Claims arrived online and on scanned paper, and specialists read and re-keyed the details, with manual look ups against other systems along the way.
Quality reviews on a small sample
Only a small share of calls could be reviewed against the quality framework, so team managers coached from a partial picture.
Tight time with operational teams
The people who know the process best were also running the day job, and system access and data protection approvals had to be worked through as the programme ran.
Deployed solutions
What We Did
A look at how we approached the work, from the first discovery sessions through to the changes we put in place, and the thinking behind each decision along the way. The aim is always the same, to deliver something measurable that the client owns and can build on, rather than a one-off fix that fades once we leave.
Guided evidence upload
An AI vision check reviews the evidence a customer uploads and tells them what is missing or wrong before they submit, so fewer applications need chasing.
Automatic application validation
Once an application is submitted, it is read, checked against the rules and routed automatically, leaving specialists to handle the exceptions.
Claim intake from any form
Claims are read from the online form or a scanned paper copy and the claim data is pre-filled for the specialist, alongside a process fix that removed a manual look up and is already live.
Call quality scoring and coaching
Every completed call transcript is scored against the company's own quality framework in one pass, with coaching for team managers linked back to the evidence in the call.
Early warning operations view
A working dashboard combines network data, weather and public signals so an emerging incident is visible before it turns into a wave of calls, parked for now to sit within the operations centre.
Real time support on live calls
A full specification and working demo for an assistant that listens alongside an adviser and brings up the right data and procedures mid call, supporting the adviser only and never speaking to the customer.