a cross-functional delivery led by the Digital Foundry
Where did it all start
The Digital Mail Centre (DMC) at Birmingham City Council (BCC) was grappling with a mammoth task: manually indexing over 2,000 emails a day across Revenues and Benefits, Council Tax, and Business Rates. This repetitive process took over 10 staff and required meticulous attention to detail to ensure documents were routed correctly in the their line of business application Acivica DM360. It’s a document management system used in BCC to store, view, and manage incoming customer and contractor documents – as well as emails and scanned post – across service areas.
The vision was clear for Sandeep Singh (Digital Mail Manager). He approached the Digital Foundry to help reduce manual effort, improve accuracy, reduce reliance on agency staff and ultimately free up his staff for higher-value work. In July 2024, we teamed up with UiPath to explore using their tool called Communications Mining (aka Comms Mining) together with Robotic Process Automation (RPA) to achieve this.
What is Comms Mining?
It’s a smart way of using technology to read and understand large volumes of messages; like emails, chats, or service requests – so we can spot patterns, and with use of RPA, can sort them automatically, and even respond without needing a person to do it manually.
It uses Natural Language Processing (NLP) to understand what each message is about – like whether someone is reporting a move, asking about a bill, or submitting a complaint. Based on what it’s trained to find, it will then categorise the message.
It can also help us understand trends in communications, identify areas for process improvement, and improve customer experience. Using Comms Mining will help us understand what people are contacting us about, this will help us identify improvement areas, reduce avoidable contact, and design better customer journeys.
Why we took this path
It gave our Developers an opportunity to learn and explore AI models – working with the DMC to train the model to recognise intent, sentiment, and extract key data from emails. One of our Foundry mantras is to ‘think big, start small, learn quickly and scale’, so we chose a phased approach, starting with emails in DM360, then scaling to scanned documents later, to build a reusable automation from the ground up.
The goal wasn’t just savings time with automation – it was about gaining further insights.
What we didn’t plan for
We started with huge excitement to try something new, and with big plans and in doing this, we missed a few things at the start:
- Stakeholder management – we didn’t fully grasp the full details of who would need to be involved and when, and the sign off we would need to get before starting was an ever-growing list – From Revenues and Benefits agreements to DTS Information Governance and Security sign off’s
- No test environment or disk space planning – our infrastructure set up took time to get signed off and put in place, environments were needed for Development, User Acceptance Testing and Production
- Delays in access for UiPath Developers – creating accounts and the right accesses to systems took more time than we anticipated
- Microsoft Exchange connection to UiPath – this integration had never been done, so was a learning curve for our Infrastructure colleagues, it also gave them an opportunity to work more closely with our Developers and UiPath colleagues
- DPIA and Security sign offs – we had to work at pace to complete a DPIA and gain information governance and security sign offs
- Resistance to ambiguity – Some of our DTS and Service teams were hesitant. Others I would describe as just unsure. This was something new and it was a bit scary! I think our colleagues were mostly worried that the technology would create problems for them further down the line. We were lucky to have experts from UiPath supporting us, and together we were able to answer stakeholder queries, and we started building and gaining trust
These gaps did slow us down, as this was our first time using this new technology and we didn’t fully grasp the necessary prerequisites. However, the bumps allowed us to pause, plan, and bring in the right people to get things moving again.
Go-live & what we gained
Together we developed an automation that eliminates manual indexing and with an AI powered solution that automatically routes emails to the correct service tray for Revenues and Benefits, allowing the DMC staff to focus time and resources on other priorities, and their customers to have their enquiries redirected more efficiently..
The first release went live in December 2024. Since then:
- 65% of emails are now auto indexed
- the backlog has reduced – 1000 emails in December 2024 compared to 5000 emails in December 2023
- team members are much happier – December 2024 was the first time Sandeep didn’t need to log in over Christmas and work and long hours!
- DMC have reported savings – 4 FTE Agency staff freed up, just in time for the annual billing drop in March this year. Had we not had the automation DMC would have been recruiting additional agency staff to cover their shortfall
- a working solution now being reused across other services
Don’t just take my word for it:
“As the automation works 24/7, we have seen a reduction in email workload, increased productivity and efficiency, fewer chase emails from citizens, reduced costs and less time spent on manual allocation. Most rewardingly, it has enabled 4 staff members to be available for training to develop themselves in other areas of DMC.” – Sandeep Singh
Lessons we learned
- Start with the foundations: Infrastructure, access, and governance matter
- Bring people with you: Change is emotional – listen, explain, and involve
- Celebrate the messy bits: Sharing our stumbles built trust and momentum
- Keep it human: This wasn’t about replacing people – it was about empowering them
- Design for reuse: The automation is now being scaled across other services, proving its value beyond the original
What are the Intelligent Automation team working on
Phase 2 with DMC is now underway. We’re expanding the AI model to be able to read scanned documents and automatically index, this is currently being investigated and could allow for further FTE savings.
Meet the makers: The teams driving innovation
From the prerequisite sign-offs to building the AI model, it was a truly a cross functional, multidisciplinary team effort:
- Revenues & Benefits services for working with us as Product Owners and allowing the team access to historic email data to help train the AI model and helped define indexing rules and exceptions
- Our Infrastructure and Applications teams – without their expert knowledge about Microsoft Exchange and DM360, we would have struggled with the integrations
- Development, Delivery and Product for working collaboratively to ensure all the technical pre-requisites were identified and met and for working with UiPath to document processes and ensure knowledge transfer
- Information Governance & Security teams ensured DPIA compliance and safe handling of citizen data
- Digital Foundry for steady leadership throughout, for navigating tricky stakeholder conversations and for keeping the team focussed
- UiPath colleagues who provided technical expertise, model training development and support, for their perseverance and in guiding us with scaling the solution in other areas.
- A super special mention to our SMEs (subject matter experts) Sandeep Singh (who also played the role of product owner), Scott Inson-Hambleton and the DMC team for trusting the process. They were pivotal in training the AI model. They worked closely with Developers to tag emails, validate predictions, and refine the model’s accuracy. Their deep knowledge of the indexing process made the AI smarter – and the automation possible.
This wasn’t just a project – it was a partnership. We built new skills, new relationships, and a new way of working. We didn’t get everything right the first time, but we learned, adapted, and delivered something we’re proud of.
If you’re thinking about automation in your own service, start small. Be honest. Share your story. And if you want to chat about how we did it – get in touch. We’re always happy to share, learn, and grow together.