Birmingham City Council: the Foundry

Birmingham City Council established the ‘Foundry’ to accelerate digital transformation and help the council close its budget gap whilst building digital skills and capability. The Foundry is helping Birmingham rethink how we deliver projects across the organisation, to a more value-driven, outcome-focused and experimental approach. As a cornerstone of the council's strategy for cultural transformation, the Foundry prioritises internal expertise over expensive external consultants. Guided by the principle of 'think big, start small, learn and scale quickly,' the Foundry fosters a collaborative environment where multi-disciplinary teams across service areas work together to streamline processes and enhance service delivery, driving efficiency, productivity, and staff engagement.

To address service area pain points, such as time-consuming manual tasks like administrative duties, redaction, translation, and mailbox indexing, the Foundry is developing and deploying AI solutions. These solutions are co-developed with officers directly involved in these challenges, resulting in more efficient, cost-effective, and fulfilling work experiences for staff.

The programme has been successful in replacing some manual processes with more efficient or automated solutions and aims to replicate the approach by scaling solutions to other services and service area challenges. This work has empowered staff to become more digitally confident and inquisitive, fostering a culture of continuous improvement driven by data-informed decision-making.

Four key back-office functions have been enhanced through the implementation of AI solutions:

• Administration: Microsoft Co-pilot has been integrated to streamline repetitive administrative tasks, such as generating summaries of legal documents and assisting with Subject Access Requests (SARs).
• Redaction: The volume of paperwork generated by disrepair and litigation claims overwhelmed the team's capacity to perform timely redactions. The development of an automated redaction tool developed using UiPath’s Document Understanding feature has enabled the team to meet deadlines and significantly reduced the risk of incurring costly fines.
• Translation: Recognising the high cost of translation services, the council developed a SharePoint-hosted tool capable of translating simple documents into 30 languages. This empowers translators to focus on more complex documents and optimise the use of direct translation services.
• Communications mining and mailbox indexing: An AI tool for communications mining and mailbox indexing reads emails to provide real-time data on citizen concerns. This allows the council to quickly address high volume areas and better meet citizen needs.  The manual sorting of thousands of emails received in a shared mailbox is a highly time-consuming process. The new AI-powered tool developed in partnership with UiPath now directs incoming emails to the revenues and benefits inbox to the appropriate inboxes, significantly improving efficiency. 

Birmingham's approach to AI implementation is guided by a detailed AI policy that has been endorsed by senior leadership. The policy recognises the introduction of new AI technologies alongside other emerging technologies, emphasising a business value-driven approach. While promoting innovation, the policy incorporates a balanced framework. This includes a pre-approved list of AI tools and stringent control measures to evaluate new AI applications against legal, ethical, security, and business suitability criteria.

A robust framework is in place to ensure the safe and responsible use of AI. This framework carefully balances the need to enhance efficiency and support new business value propositions while effectively mitigating risks.

Staff engagement and collaboration are central to the Foundry's approach. Solutions are developed collaboratively in a problem-centred manner, ensuring they meet the needs of end-users. Working in 6-12 week sprints, the team begins by engaging with staff to identify service delivery areas ripe for improvement and efficiency gains. Through these discussions, they gain a deep understanding of staff challenges, existing processes, and the level of team support for potential solutions.

The development process is iterative, with continuous feedback from users informing the ongoing refinement of tools. Collaboration extends beyond the initial challenge-setting phase, with Foundry teams and service area teams co-designing solutions. For instance, when developing the redaction solution, officers responsible for manual redaction worked closely with software developers to train the AI model and identify other workflow complexities. This collaborative approach has fostered strong buy-in among officers, empowering them and instilling a sense of ownership in the implemented solutions.

Beyond solution development, this approach champions officer development by exposing them to new ways of working and enhancing their digital skills. This fosters a broader mindset across the organisation, encouraging greater adoption of innovative approaches.

While still in the early stages of development, the AI solutions implemented by the Foundry are already demonstrating positive outcomes across all four areas. By streamlining repetitive administrative tasks, these solutions have generated significant efficiencies, resulting in both time and cost savings for the council.

Benefits across each focus area:

• Copilot: Automating tasks like Subject Access Requests (SARs) has increased efficiency and freed up officer time for more meaningful work, leading to increased job satisfaction. 
• Translation: AI-powered translation has the potential to generate substantial cost savings once fully embedded and adopted.  Microsoft’s AI translation tool was easy to deploy and was instrumental in the early days of Foundry to showcase how internet age technology can transform how we work. Birmingham previously spent nearly £350,000 on translation across the council per year, a cost that is expected to reduce with the new technology. This allows for the prioritisation of translation resources for complex documents and in-person interactions. 
• Redaction: The Foundry has adopted a phased approach to the redaction challenge. The first phase focused on developing and training an AI model to automate the redaction of personally identifiable information, freeing officers from time-consuming manual tasks. The redaction tool has been deployed for 3 document types so far. This has a 95 per cent redaction success rate and is redacting automatically in 90 seconds compared to 10 mins of human effort. The ultimate goal is to implement a comprehensive end-to-end solution that incorporates Robotic Process Automation (RPA) to further reduce manual work. 
• Communications mining & mailbox indexing: The AI-powered mailbox indexing bot is successfully indexing 55 per cent of emails, resulting in an annual saving of £100,000 and freeing up over 5,500 staff hours. We are continuing to work with the team to improve the success rate and free up even more time.

Beyond efficiency gains and cost savings, the Foundry's approach and the implementation of AI tools have fostered a positive shift in the council's culture. By reducing the burden of repetitive tasks, officers can now dedicate more time to meaningful and value-added work, significantly enhancing workforce engagement.

Furthermore, the Foundry has played a crucial role in cultivating a more technology-driven culture within the council. By actively involving officers in the development and implementation of AI solutions, even those with limited prior technology experience have gained valuable skills and increased confidence in leveraging technology within their roles. This collaborative approach has not only transformed the work culture within some teams but has also fostered a broader sense of openness and curiosity towards the potential benefits of technology across the council.

While the implementation of AI has been successful improving some business functions in Birmingham, it has also generated valuable insights and learning for future AI development and implementation.

Scepticism about AI generated outputs

When notes were generated by AI instead of a human, the notes faced an increased level of scrutiny despite being of an equivalent quality. 

Limitations with AI translation

The initial implementation of AI for translation demonstrated limitations, highlighting areas where the technology needs further development. While this resulted in less impact than initially hoped, it provided valuable insights for future rollouts. Moving forward, a key consideration for successful AI deployments is a structured evaluation of the suitability of the intended use case. For example, the translation tool proved highly effective for straightforward documents across multiple languages, however, its application in face-to-face interactions or with lengthy complex text presented challenges.

Built-in guardrails

Some AI platforms incorporate built-in safeguards that restrict the input of information deemed inappropriate or overly sensitive. These safeguards have presented challenges for certain services, such as those working with children and families, who routinely handle sensitive data. These services have found it difficult to utilise certain AI tools due to these restrictions.

These challenges and considerations have underscored the importance of carefully considering the limitations of AI implementation alongside its potential benefits. Furthermore, the experiences with initial AI deployment have highlighted the value of learning from both the successes and the challenges in navigating functionality and risks associated with AI adoption.

Building upon the success of the Foundry and the collaborative approach to digital transformation within the council, Birmingham aims to leverage AI to drive innovation and enhance operational efficiency. This aligns with the council's broader strategy of fostering a culture of continuous improvement. To achieve this, immediate priorities include establishing robust data management practices and continuing the strategic implementation of AI tools in critical areas such as translation and meeting minutes.  The Council has mobilised a stable multi-disciplinary team for Intelligent Automation initiatives within the Foundry. This team is currently developing ways of working, standards, principles and their delivery approach. The team is looking to continue to scale the use of the capabilities developed, such as redaction and communications mining.  The team will utilise the learning the team has gained to improve the management information and data we have about current service provision and failure demand to expose valuable problems to solve, and solve common problems at pace.