Resetting the relationship between local and national government. Read our Local Government White Paper
Welcome to the LGA AI use case bank, which has been compiled to inform and empower local councils in their exploration of Artificial Intelligence (AI) solutions.
Building on the findings of our recent AI survey, which highlighted the need for practical examples, this bank showcases the diverse ways local authorities are leveraging AI.
Within this collection, you'll discover a spectrum of AI adoption, ranging from utilising AI assistants to streamline back-office tasks to pioneering the implementation of bespoke Large Language Models (LLMs). These real-world use cases exemplify the innovative spirit driving advancements in local government service delivery.
Whether your council is at the outset of its AI exploration or seeking to expand its existing capabilities, this bank offers a wealth of valuable insights and best practices to support your organisation's AI journey.
Overview
Barnsley Council is emerging as a frontrunner in local government's adoption of artificial intelligence (AI). Through a strategic partnership with Microsoft, they've implemented Microsoft Copilot on a large scale. This innovative approach aims to utilise AI to automate administrative tasks, boosting efficiency and empowering staff across various departments.
Recognising the potential of Copilot early on, Barnsley secured a place in Microsoft's early access program (EAP) back in September 2023. Building a strong business case that quantified the benefits, the council took a decisive step to expand its pilot program significantly, from 300 to 2,000 licenses. This ambitious rollout ensures the majority of their staff have access to Copilot, aiming to streamline workflows, improve efficiency, and free up valuable time previously spent on repetitive tasks.
The council were able to roll out the licences within a week, starting in late December and issuing licences across the organisation by the first week of January 2024.
Engagement and support
Barnsley Council's commitment to stakeholder engagement was evident from the start of the adoption process. Early involvement of elected members, trade unions, and senior management teams ensured alignment and addressed potential concerns. Demonstrations of copilot played a crucial role, not only showcasing technology's capabilities but also developing a shared understanding and organisational appetite for AI deployment across all levels.
Building on this strong foundation, the council implemented a comprehensive support programme to empower its workforce. Regular interactive workshops equipped staff with the skills to utilise Copilot effectively. The creation of a dedicated "Copilot Flight Crew" – a community of champions – fostered a collaborative environment. This champion network thrives on a central team site, where members share prompts, success stories, and best practices. The site also acts as a hub for ongoing training, ensuring staff continuously develop their AI literacy and refine their writing skills to maximise the benefits of Copilot.
This comprehensive approach extends beyond the initial training. Over 150 staff members were trained as champions, solidifying their expertise and enabling them to provide ongoing guidance and upskilling to their colleagues. This peer-to-peer support network fosters a sense of community and ensures the knowledge gained from Copilot implementation is readily shared and sustained.
AI Governance
Barnsley Council prioritises the responsible use of AI through a comprehensive governance framework. This framework comprises:
- An ethics board: This board, composed of social care practitioners, housing practitioners, and technical staff, tackles high-level issues such as consent and data security.
- Enhanced information governance: An addendum to the existing policy clarifies that all AI-generated content must be vetted by humans before use.
- User training and workshops: Ongoing training reinforces the proper use of AI and ensures users understand their responsibility for AI outputs.
- Risk-specific protocols: In "high-risk areas" like children's social care, the council implements specific rules for data handling within Copilot.
These measures aim to build trust and transparency with stakeholders. They demonstrate the council's commitment to ethical AI implementation and continued human oversight.
Benefits and impact
Barnsley Council's early investment in AI has yielded significant results. From the outset, they identified finance, business support, and social care as areas ripe for improvement. Copilot's impact has been substantial:
- Staff across departments report time savings in tasks such as meeting transcriptions and minutes, document summarisation, and administrative support. Additionally, analysing Copilot transcribed meetings has helped identify areas for improved meeting efficiency and chairing techniques.
- Procurement processes benefit from Copilot's ability to answer qualifying questions, while legal services leverage document summarisation to focus on complex tasks. Finance teams have seen improvements in spreadsheet usage and macro debugging.
- Social care workers experience a significant reduction in administrative burden, freeing them to dedicate more time to direct client care.
- Usage data shows a strong 70 per cent of users regularly engage with Copilot, with many staff keen to retain their Copilot licence, highlighting its value.
Overall, the council is experiencing substantial efficiency and productivity gains, especially in areas previously bogged down by administrative tasks.
Looking ahead, the council actively measures the return on investment (ROI) by assessing productivity and cost savings. Their goal is for Copilot to generate significant cost savings, exceeding its initial investment. Specific metrics include reductions in agency staffing costs, efficiency gains in administrative tasks, and improved productivity across departments. Based on Copilot's initial success, the council is committed to renewing licences and proving the tool's ROI within the next year.
What next
Barnsley is just at the beginning of their AI journey. The successful pilot of Copilot has ignited enthusiasm for AI's potential to transform service delivery. The next phase focuses on deepening the impact of Copilot. This will involve revisiting specific service use cases to refine how they measure return on investment (ROI) and build a compelling business case that secures continued investment.
Alongside this, the council plans to expand Copilot adoption across the organisation. Barnsley will provide further support to staff to ensure a smooth integration and knowledge of copilot is developed throughout all service areas. Additionally, Barnsley seeks to unlock Copilot's full potential by exploring advanced integration with the wider Microsoft suite. This includes leveraging connectors and data lakes to empower managers with direct data querying through Copilot and Power BI. Additionally, integrating Copilot with HR policies could streamline staff access to information, potentially replacing complex intranet structures. Complexities exist with linking up technology outside of the Microsoft suite, but the council are actively exploring solutions.
Buckinghamshire Council stands out as a pioneer in adopting Microsoft Copilot. They were among the first councils to embrace this AI technology, granting them valuable experience and a chance to shape the technology's development. This case study has been collected via in-depth discussion with Tony Ellis (IT Director) and Peter Parfitt (Head of Digital and AI Programme Lead) who have led and overseen the council’s early adoption of Copilot. This case study explores Buckinghamshire's journey with Copilot, highlighting the positive outcomes, challenges encountered, and lessons learned.
Early Adoption and Strategic Rollout
As a Copilot early adopter, Buckinghamshire had access to the AI tool in October 2023 before the technology had been generally released. The council strategically rolled out 300 Copilot licenses across various departments. Initially, the council set five key strategic objectives for the pilot of the technology:
- Enhanced staff productivity
- Improved work quality
- Achieved cost savings
- Ensured risk aversion
- Promoted staff well-being
Early adoption offered Buckinghamshire distinct advantages. Microsoft provided dedicated support, including training sessions that proved highly popular – attracting over 140 attendees despite being held on a Friday afternoon. However, being among the first councils to trial Copilot also presented challenges. As the council was at the forefront of implementation, some aspects of the technology as well as the associated advice and training were still under development and continue to be developed.
To fuel interest in their copilot trial, Buckinghamshire hosted an innovative "Dragon's Den" style event. Teams from nine different departments (including HR, IT and finance) pitched how Copilot could benefit and improve their work. This not only identified several promising use cases but also fostered excitement and buy-in across the organisation.
AI Governance: A proactive approach
In a pre-emptive move, Buckinghamshire established an AI governance board in October 2023, before deploying Microsoft Copilot. This ensured senior management and members were engaged with AI initiatives from the outset. Chaired by a corporate director, the board oversees all AI endeavours, fostering transparency with residents and partners. It emphasises that AI acts as a supporting tool, not a final decision-maker.
The board leverages technical expertise across the organisation, including information governance colleagues. The board convenes as regularly as every fortnight, highlighting the speed of adoption within the council. Regular meetings showcase progress with Microsoft Copilot, highlighting successful use cases. Discussions delve into ongoing challenges associated with AI adoption, alongside a dedicated risk register to guarantee proper risk management and mitigation. Essential processes and procedures, including usage policies, Data Protection Impact Assessments (DPIAs), and Equality Impact Assessments (EQIAs), have also been implemented by the board to ensure safe AI deployment.
The AI governance board actively manages Copilot-related risks through a dedicated risk register process, incorporating Microsoft's six principles for responsible AI. Regular reviews and mitigation strategies ensure potential issues are monitored and addressed. Enhanced confidentiality agreements with specific staff were implemented to bolster data security. Clear communication emphasises that Copilot assists with tasks; final decisions and accountability remain human-centric. An attribution statement is currently being developed to clarify AI co-authorship within reports.
Training and support: Building capabilities
Buckinghamshire provided comprehensive support to staff. AI leads offered one-on-one training sessions on copilot’s application to existing systems, a champions scheme was established, and regular demonstrations showcased Copilot's use cases within different departments across the council. As formal Microsoft training wasn't available initially, Buckinghamshire fostered peer-to-peer learning, crucial for the initial rollout. The council are now refining training based on experience to prepare for wider adoption.
Measurable benefits: Time savings and beyond
Buckinghamshire prioritised measuring the benefits of Copilot. Following the initial rollout, they focused on quantifying the return on investment (ROI) to align with their financial plans. From the initial pilot of the technology, the council have seen several tangible benefits of utilising Copilot in key service areas.
- Customer Service: Copilot is being utilised to transcribe and summarise calls, saving agents an average of 2 minutes per call wrap, equating to a 33 per cent reduction. This is significant considering there are more than 100 customer service agents who receive a high volume of calls every day.
- Executive Assistants: Assistants using Copilot have reported a 20 per cent reduction in task completion time (60-90 minutes saved daily) for activities such as creating reports and meeting minutes. This data was capture from closely monitoring a group of 6 executive assistants.
- Project Management: Copilot streamlines project initiation by capturing meeting transcripts and prompting population of project documentation templates within the IT department.
- Human Resources: HR uses Copilot for sentiment analysis and job description/interview question writing. Copilot helps compare interview transcripts to job specs and pre-written questions, reducing advert creation time from hours to minutes.
- Children's Social Care: Copilot is being piloted to condense information from multiple documents into standardised education, health, and care plans.
An unexpected benefit of deploying Copilot in Buckinghamshire has been its positive impact on colleagues with neurodiversity. For example, the advanced transcription capabilities have significantly transformed the daily work of a deaf colleague. Unlike subtitles, transcription software offers a more comprehensive picture of conversations, allowing deaf participants to identify speakers, capture action items, and track unresolved issues. This has been particularly transformative, enabling the colleague to fully participate in meetings, review discussions later, and independently verify emails for accuracy, reducing reliance on others.
Challenges and considerations
Despite identifying numerous benefits, Buckinghamshire also encountered limitations and challenges.
- Adult Social Care: Initially, Copilot exhibited issues with "hallucinating" or fabricating information about patients. The council collaborated with Microsoft to develop specific prompts that addressed this problem, ensuring data fidelity.
- Inconsistency: The same question might yield different responses. Staff are adapting by understanding Copilot as a tool for generating helpful suggestions, not guaranteed consistency or accuracy.
- Data and Functionality Limitations: Copilot currently struggles with quantitative data and isn't fully integrated with Microsoft Excel. However, ongoing development suggests these limitations may be addressed in future updates.
Looking ahead: Building on the foundation
Buckinghamshire sees immense potential in Copilot beyond its current applications and are excited to continue their adoption journey as the technology develops. They are actively exploring new use cases as they continue their pilot.
To secure continued investment, the council is developing a more sophisticated method for calculating ROI to strengthen their business case before license renewal in November 2024. While quantifying the value of improved staff well-being can be challenging, the council recognises the importance of a holistic view of Copilot's impact.
In conclusion, Buckinghamshire’s pioneering approach to Microsoft Copilot adoption serves as a valuable case study. Their proactive measures in governance, training, and exploration provide valuable insight for councils looking to adopt Microsoft Copilot.
Quote from Deputy Cabinet Member for Resources, Councillor Jonathan Waters who said:
“It is early days, but we are already seeing real results in areas like our customer contact centre, which deals with around 8,500 calls a week. Using AI is improving quality and complaints procedures, as well as streamlining response drafting and call summarising. We believe even bigger wins will follow when we begin to employ Copilot to enhance frontline services, such as social care and housing, which take up the bulk of our budget. If we can be more efficient, that's fantastic and if we can trim just 1 per cent off the cost of social care, that's where the real pressures are.
As AI continues to expand, it is vital that it is deployed ethically, and we have a dedicated AI Governance Board overseeing its use across council services. This is especially important in the public sector. Using AI promises a future where we can proactively identify areas for improvement before they become a drain on resources. I’m confident we will continue to see the benefits."
Overview
Halton Council embarked on an ambitious project to leverage AI to enhance their support offer to families. The initiative aims to improve access to services, streamline processes, and provide personalised support to families. Halton Borough Council, with funding from the Department for Education and the Department of Health and Social Care, has created a Digital Front Door into its Family Hub Services partnering with Beebot AI. The Family Hub is in the implementation stage and is currently in the process of being embedded within the council.
The Family Hub is an interactive desktop and mobile application that offers 24/7 access to children’s and families’ services, addressing the challenge of traditional service hours that may not suit many families' schedules. Automated processes streamline tasks such as appointment scheduling and information dissemination, freeing up staff to focus on more complex issues. AI-driven chatbots are available to answer common questions and provide guidance on a range of topics.
The platform uses AI and intelligent automation to deliver tailored information and support to families based on their specific needs and preferences. Personalisation is available through the Family Hub, which allows individuals to sign up for push notifications relevant to their current situation. These push notifications will send relevant and timely information directly to individuals, such as updates on the different stages of pregnancy or on children's immunisations, which can then act as signposts to further information and other services.
The platform provides detailed analytics, offering the council valuable insights into user behaviour. As more residents utilise the app, data driven decision making becomes increasingly possible, enabling the council to better tailor support to residents’ specific needs and make iterative improvements to the app. Should residents require further assistance, they can easily contact an officer through the live chat function. Automated processes have begun to streamline workflows and reduce the administrative burden on staff.
Governance and stakeholder engagement
The Family Hub is governed by a monthly steering group, which has been instrumental in the initial discussions about AI implementation and kept senior staff informed and engaged. Discussions have centred on the financial sustainability and potential benefits of AI in service delivery. To effectively communicate the benefits of AI, tangible demonstrations and a proof of concept were crucial, as abstract concepts are often difficult to grasp. Initial concerns about AI were addressed through explanations of its role in complementing existing services and freeing up staff time. To build further support, demonstrations of the AI platform were presented to stakeholders, which helped to generate buy-in.
Given the complexity of children’s and families’ services, initial engagement focused on internal stakeholders, such as service providers and organisations. Once their support was secured, the focus shifted to engaging service users.
To ensure a comprehensive approach, digital working groups were established with key experts to discuss content development and future plans. Additionally, a weekly digital projects group meets with Beebot AI to maintain momentum and address ongoing implementation challenges. While the governance process is in place, the scale of the project and resource constraints may sometimes limit the speed of progress.
Risk management
Considerations for the safety of users of the Family Hub are being made as the app continues to develop and embed. Halton are working on the app to ensure that the use of the app will not endanger any users. Potential risks, such as privacy concerns for domestic abuse victims, have been identified and discussed. To mitigate these risks, the platform is being designed to minimise digital footprints and ensure the safety of vulnerable users.
The platform is designed to ensure that search results are accurate and reliable. By restricting the AI to information inputted into the app, the Family Hub safeguards against generating misleading or inaccurate advice.
Delivery strategy
The initial phase of the rollout involved a two-pronged approach: first, engaging professionals through a launch event and ongoing training to ensure they understood the platform's benefits and could effectively communicate them to families. A launch event in July 2024 attracted 140 professionals, who were crucial in embedding the technology as standard practice within their work with children and families.
Secondly, a communication campaign was launched to directly target families, using various channels such as social media and digital animations to explain the platform's features and encourage usage.
The long-term vision for the Family Hub is to create a single digital front door for all family services, streamlining access to information and support. AI will play a crucial role in automating processes, such as event booking, and improving efficiency. By increasing accessibility and reducing administrative burdens, the council aims to benefit both families and staff.
Increasing accessibility of services
The Family Hub offers increased accessibility by allowing users to access information without direct interaction with an officer. This is particularly beneficial for groups who may prefer digital engagement, such as fathers who sometimes feel excluded from traditional family services. By providing a digital platform, the Family Hub empowers residents to access information and support without the need for face-to-face interactions. Additionally, the council is committed to making the platform accessible to all users, including those who are deaf. Efforts are underway to develop features that cater to the needs of the Deaf community, such as sign language support.
Additionally, Halton is working with other councils in the North West to identify challenges and share best practice. By collaborating with fellow councils, Halton aims to address potential challenges and improve the overall accessibility of the platform.
Next steps
The engagement phase which is targeted at families is underway, with the aim of increasing adoption and use of the app amongst residents. Halton's approach is centred around strong collaboration with internal staff, partners, and other councils. By maintaining open dialogue with these key stakeholders, the council continues to optimise the effectiveness of the app.
The council plans to further expand the use of AI within the Family Hub programme and explore opportunities to leverage AI in other areas across children’s services. By embracing AI, Halton Council is enhancing its digital capabilities and making key services more accessible to those who need them most.
Overview
After four years of collaboration, Hertfordshire County Council, technology company Robotiz3d and academics at the University of Liverpool have successfully developed an innovative AI-powered solution for road maintenance. This partnership combines Robotiz3d's expertise in AI and robotics with the Council's knowledge of managing highway assets. Their collaborative goal was to create an innovative system for identifying and repairing road defects, focussing on cracks that become potholes.
The project stemmed from Robotiz3d's initial technology, ARRES (Autonomous Road Repair System) Eye, used for road scanning. This technology is used within the ARRES Prevent robot, initially backed, and funded by Innovate UK. The robotic technology can autonomously detect and fix road cracks, preventing them from becoming potholes.
The robot uses AI to pinpoint and assess cracks, potholes and other road imperfections. It then automatically fills cracks to prevent water from seeping in and forming potholes. More than 75 per cent of cracks in the road surface become potholes within three years, so they are a major problem for the whole sector.
The pilot project aimed to achieve two things: Firstly, to conduct the world's first public road trial of the technology in Hertfordshire. Secondly, to garner positive publicity and showcase the Council's commitment to innovation and proactive solutions for potholes and road maintenance, a major concern for residents.
Robotiz3d developed the proprietary ARRES Eye scanning technology and the prototype ARRES Prevent robot. The AI model powering this technology was trained on vast datasets, including detailed scans of Hertfordshire's road network. The council collaborated with Robotiz3d to improve the system's accuracy. The project team scanned various roads around Hertfordshire, which significantly increased the technology’s pothole identification accuracy from 66 per cent to 100 per cent. The model also learned to recognise new road features and defects it hadn't encountered before.
To optimise the AI model and robot's functionality for real-world use, Hertfordshire County Council closely collaborated with Robotiz3d. The Council provided valuable data from their road network and expertise from their highways asset management professionals. This joint effort allowed for iterative refinements to the AI model and robot based on the team's feedback and insights. This ensured the AI-powered robot was fully prepared for real-world testing.
Engagement
The Council involved key stakeholders early on to facilitate real-world testing and ensure buy-in. Ringway, the Council's maintenance contractor, played a critical role by providing expertise on health and safety at the test sites. Elected members were also informed and engaged at the project's outset, fostering awareness about the Council's use of AI and generating enthusiasm among staff. Their positive feedback further bolstered the project's momentum.
The Council's communications team took a proactive approach as the project progressed, securing positive media coverage. This included a feature on BBC One's The One Show, which significantly increased public interest, both locally and nationally. Council members were also given opportunities to speak on various news channels, highlighting Hertfordshire's innovative approach to road maintenance and the potential benefits for drivers. The communications team actively monitored and responded to public queries and comments on social media, where organic posts reached 125,000 people, with over 31,000 views of a Tik Tok video, maintaining transparent and timely communication with the community.
Pilot impact and global press coverage
The Hertfordshire County Council and Robotiz3d partnership has fostered mutual learning throughout the project. Robotiz3d gained crucial real-world insights. This includes the logistics of robotics deployment (transport, storage) and integrating with existing maintenance practices. Hertfordshire, in turn, shaped the pioneering technology by offering asset management expertise and feedback to refine AI models and robot functions. This "learn by doing" approach tackles the challenges of implementing cutting-edge tech in complex environments. The initial pilot has fixed a small number of cracks and Hertfordshire engineers will be reviewing the long-term benefits of the repair.
The project's initial impact is undeniable. The world-first trial announcement garnered significant media attention, with over 100 outlets reaching an estimated audience of 2 billion globally. This showcased the Council's commitment to innovation and technological advancement for its residents.
Beyond PR, this AI technology holds immense potential for cost and time savings in road maintenance. By enabling early intervention on cracks before they worsen, the technology prevents the need for expensive repairs later. This proactive approach has the potential to significantly reduce lifecycle costs associated with road infrastructure.
Risk mitigation and negation
While personal data isn't involved, governance is still important for managing risks to workers and managing road closures during the operation. Hertfordshire County Council and Ringway leverages their existing, robust site risk assessments for highway work. They are in the process of developing a specific protocol for the robot's safe interaction with human crews.
The inherent risks of roadwork necessitate a focus on safety. This AI-powered robot has the potential to significantly reduce worker exposure to danger while supporting a critical council service.
Resident and road user safety remains paramount throughout the pilot. The Council's existing risk assessments, though rigorous, are constantly reviewed. As the robot's use evolves, potential new risks will be identified, and the assessments will be adapted accordingly.
What next?
The Hertfordshire trials hold promise for addressing the UK's staggering £16 billion road repair backlog. This AI-powered robotic technology offers a potential solution to bridge the gap in road maintenance. By successfully demonstrating its effectiveness, Hertfordshire paves the way for wider adoption across the country.
The project's true return on investment (ROI) will unfold over time. As the technology evolves based on Hertfordshire's lessons learned, cost savings and operational benefits are expected to multiply. The AI will become more sophisticated and efficient, further enhancing its value.
Hertfordshire plans further tests to refine robot usage and develop county-wide deployment strategies. This will ultimately lead to significant improvements in road quality across the region.
Kent County Council shared their approach at the Local Government AI Network.
Overview
To reduce the time burden administrative tasks place on social workers, Kingston Council has been working collaboratively with Beam to pilot Magic Notes. This AI-powered solution aims to streamline the process of writing case notes and assessments, freeing up valuable time for social workers to focus on direct client care.
The demand for adult social care services has been steadily increasing across most local authorities, posing a significant challenge to the sustainability of existing care frameworks. According to the NHS’s Adult Social Care dashboard, there has been an average annual increase in new case demand of 2 to 5 per cent since 2017, with Kingston experiencing a surge of approximately 14 per cent. Traditionally, social workers take minimal notes during care visits to maintain personal, face-to-face interaction with clients, but this can mean it takes longer to complete a report afterwards. By using Beam to automate the transcription and summarisation of care visits and to output those summaries in a format that can be easily input to the case management system, Kingston seeks to make a substantial impact not just on the efficiency but also the effectiveness of adult social care delivery.
The project aimed to adapt Beam's existing product, which was initially designed for their own caseworkers, to be used by social workers on mobile ‘phones during real-world conversations. The primary objective was to create a simple and user-friendly solution, accounting for the varying technical proficiency of social workers. The development process involved creating new user journeys to ensure the solution was intuitive and easy to use on mobile devices.
To train the AI model, anonymised data from templates and assessments previously written by social workers was provided to Beam’s AI. This data was used to ensure that the AI's output matched the required level of detail and populated the correct fields within the case management system.
The development process was completed at pace. Launched in January, the solution was ready for pilot by the start of February. The pilot initially included only three social workers, but quickly expanded to encompass over 10. Use cases grew from the initial case reviews only to include supervisions, occupational therapy assessments, and care needs assessments, enhancing the tool's functionality and value.
AI governance
Kingston has made keeping our client’s data safe a primary principle of the project. As a result, Beam deployed an enterprise version of ChatGPT within their environment, ensuring all sensitive data remained under their control. A comprehensive Data Processing Agreement (DPA) was established with Beam, outlining data access, processing protocols, and employee Disclosure and Barring Service (DBS) checks.
Furthermore, a thorough Data Protection Impact Assessment (DPIA) proactively identified and mitigated potential data privacy concerns. By meticulously examining data flows, storage mechanisms, and access controls, the council pinpointed and addressed potential risks.
As a result have been made to Kingston’s online privacy notice to encompass Magic Notes usage, but practitioners will be expected to obtain consent for recording from meeting participants and explain the purpose of it to service users.
The rollout plan, along with full-scale implementation details, has been approved by Kingston’s Adult Social Care Transformation Board and a staged implementation is now under way.
Although the project is now live, we haven’t finished working with Beam to improve the system by continuing to test new templates, gathering user feedback, and making necessary adjustments to ensure Magic Notes remains effective and compliant. Beam and Kingston have developed a feedback process for staff to document how much time they’re saving and, most importantly, what they are managing to achieve with that time. Kingston receives a weekly digest of the feedback; it helps Kingston and Beam tune the templates as well as highlighting areas where training or further engagement may be necessary to improve takeup.
Engagement and staff support
Kingston Council prioritised stakeholder engagement throughout the Magic Notes pilot. Social workers and the Adult Social Care operations manager witnessed a live demonstration to assess the tool's potential. Beam and Kingston’s digital team then collaborated with participants to define use cases for the proof of concept. Regular feedback sessions with participants and the manager ensured continued progress and addressed any concerns. To gain leadership buy-in, the heads of Adult Social Care and Digital & Data were also kept informed.
The training was designed to empower social workers for successful Magic Notes adoption. The comprehensive sessions covered project goals, benefits, and impact on daily workflows. Hands-on training familiarised participants with the application, its key features, functionalities, and best practices. Social workers explored the app
in a controlled environment, practising logging in, navigating, and using features. Role-playing exercises with colleagues further solidified learning by simulating real-world scenarios with Magic Notes. Finally, Beam offered shadowing support for additional on-site guidance.
Initial impact: Saving valuable time for social workers
Initial results from the Magic Notes pilot are promising. Social workers reported significant time savings in completing case notes and assessments, with an average reduction of 50 to 60 per cent. The most dramatic impact was observed in supervision write-ups, where the process went from taking 40 minutes to less than 10 minutes
using Magic Notes.
The accuracy of Magic Notes' transcripts surpassed expectations. While the project team initially aimed for 85 per cent accuracy, the results yielded an impressive 96 per cent plus. This efficiency extends beyond on-site visits, as social workers have utilised the tool to dictate notes for AI transcription and summarisation.
Social workers highlighted that they regularly alter the transcriptions and summaries to tailor them to their own writing style and add in additional detail, emphasising Magic Notes as an assistive tool rather than a replacement for their expertise.
Kingston Council employed a data-driven approach to assess Magic Notes' financial viability. Focusing on time savings demonstrated, the council have been able to provide an efficiency saving to improve our prevention services.
To further strengthen the case, the council analysed specific use cases like supervisions, where Magic Notes reduced note taking by around 75 per cent. They calculated the net financial benefit per supervision, factoring in both time saved and the Magic Notes solution cost.
Recognising the tool's scalability, the council could potentially reduce their cost of time writing assessment and case notes by up to 50 per cent.
Challenges and limitations
The evaluation identified limitations for service users with low verbal skills or mental capacity. Magic Notes relies on clear audio, and conversations with reduced detail or poor audio quality resulted in inaccurate transcripts. This shows that Magic Notes is not a perfect solution for all user groups. Testing continues with these user groups
though to see if the product can be improved.
Technical challenges emerged during the initial proof of concept, including name misrecognition, information repetition, and background noise impacting accuracy. However, the strong partnership between Kingston Council and Beam ensured a swift resolution. Beam introduced features to reduce the impact of background noise and improve accuracy and also provided feedback mechanisms for social workers to
refine templates, fostering continuous improvement.
One thing that came out during testing was different use cases for what was initially thought of as a single process - supervision. Managers supervising seniors or team leaders talked about different things to seniors supervising case workers and so a different template was required.
The iterative improvement of the tool driven by Kingston social workers has paved the way for several other councils to pilot Magic Notes.
Contact [email protected] if you would like to find out more.
Overview
Newham Council, a forward-thinking local authority, recognises the potential of AI to improve its service delivery. The University of East London (UEL), a local institution with a strong track record in AI research, shares this vision.
The organisations have a history of collaboration together and have forged a strategic partnership to establish a UK-based centre dedicated to pioneering AI research, training and solutions for the public sector. By applying AI to tackle complex local government challenges, the centre aims to drive innovation and deliver technology-enhanced solutions that meet the community's needs. With a broad scope encompassing key service areas, this collaboration is poised to offer valuable insights as other councils embark on their AI journeys.
The ambitious partnership aims to:
- To promote the use of AI in improving services in areas such as temporary accommodation, local economy, social care, and climate change mitigation.
- To ensure ethical and effective use of AI through robust governance and compliance frameworks.
- To contribute significantly to the Sustainable Development Goals (SDGs), particularly in areas such as Arts and Culture, Young People, Sustainability, Health Inequalities, Economic Inequalities, and Data Economy or Cybersecurity.
Delivery strategy and governance
Building on the Newham Data Analytics Framework (NDAF), the joint venture will adopt a comprehensive approach. This involves establishing a robust system for managing data from various sources, conducting research and development to uncover new insights, particularly in areas like health and education, and ensuring strict governance and ethical practices for AI. The project will also prioritise community involvement and collaboration across different academic disciplines. To sustain progress, training and capacity building initiatives will be implemented for both the public sector and local community.
To ensure strategic direction, the joint venture will be led by a team representing Newham Council, the University of East London, and the London Office of Innovation and Technology (LOTI). An advisory board consisting of experts from academia, industry, and the public sector will provide additional guidance. This collaborative approach aims to produce research findings applicable to the wider local public sector.
A strong emphasis will be placed on the ethical use of AI through robust governance frameworks. The partnership is committed to aligning its work with the Sustainable Development Goals, focusing on areas such as health, economic inequality, and youth engagement.
Next steps
The next steps involve engaging with stakeholders to refine project objectives and securing funding to establish the venture. Pilot projects will be launched in key areas to demonstrate AI's potential, followed by training programs for public sector staff. Continuous community involvement is essential to align projects with public needs, while partnerships with corporations and government bodies will ensure ongoing support and collaboration.
The initial focus of the partnership will be to address the pressing issue of temporary accommodation in Newham. With the council's expenditure on temporary housing exceeding £20 million in the past year, one of the highest spends in the country. This area presents a significant opportunity for AI-driven solutions. Drawing on existing data, the partnership aims to develop a predictive model to forecast housing demand and homelessness trends. Additionally, an AI system will be considered to streamline the procurement and allocation of suitable temporary accommodation properties.
Conclusion
The LGA eagerly anticipates the Centre for AI's development and the potential benefits it will bring to the wider public sector. By concentrating on critical local government challenges and fostering collaboration between academia, industry, and the community, this partnership aims to deliver substantial improvements in service delivery, policy formation, and overall community outcomes. The Centre's work has the potential to be instrumental in shaping the future of AI in local government.
Overview: A proof of concept journey
Children's social workers are the linchpin of child protection, yet information overload can hinder their effectiveness. North Yorkshire Council identified a critical challenge: social workers dedicate 80 per cent of their time to administrative tasks due to the sheer volume of data they manage. This includes case notes, forms, assessments, and uploaded images, all stored in disparate formats within the case management system (Liquid Logic LCS).
This data sprawl creates two major problems:
- Time-consuming searches: Finding the information needed often requires trawling through multiple sources, delaying timely intervention for vulnerable children.
- Limited access to historical data: Crucial insights can be buried within years of casework data, making it difficult to draw on past experiences to improve outcomes.
Recognising these limitations, the council embarked on a bold initiative. Funded by the Department for Education's Data and Digital Solutions Fund, they developed a proof-of-concept project to reimagine case management using Artificial Intelligence (AI) powered by Microsoft Azure.
The proof of concept project aimed to:
- Reduce social worker administrative burden through the development of a secure and scalable AI-powered infrastructure on Microsoft Azure.
- Improve information retrieval efficiency by enabling advanced semantic search capabilities that can query both structured and unstructured data within the existing case management system (Liquid Logic LCS).
- Enhance decision-making with data insights by creating auto-generated eco-maps that visualise the networks and relationships surrounding children and families.
- Develop knowledge-sharing resources by creating comprehensive practice guidance and transferable "How-to" guides for other local authorities or organisations.
Strong stakeholder engagement and governance
North Yorkshire Council ensured project success through robust stakeholder engagement and governance. This included:
- A diverse project team and board, with members from various services and disciplines provided comprehensive expertise.
- Regular communication with stakeholders including open webinars, Department for Education events, and local events facilitated information flow.
- The project adhered to the council’s existing transformation governance structure, reporting to relevant boards.
- An AI Steering Group was created, the high-level group (Director and Assistant Director level) provided strategic oversight and aligned the project with the council's AI strategy.
- An organisational communication plan ensured everyone at North Yorkshire understood the project's goals and implications.
Additionally, North Yorkshire employed a robust risk management approach aligned with its operating procedures. The project team and board maintained a centralised risk and issue log, outlining clear mitigation strategies and escalation processes for identified risks. Risk workshops were conducted to support in-depth analysis, and this process ran concurrently with the Data Protection Impact Assessment (DPIA), Equality Impact Assessment (EQIA), and Ethical Impact Assessment (EIA).
Ethical considerations at the forefront
Due to the data's sensitivity, the project prioritised ethical considerations. A comprehensive ethical impact assessment, based on the Socitm’s Ethical Framework, explored potential risks and benefits. The council engaged key stakeholders in various discussions around the ethical considerations of the project. Key ethical considerations shaping the project's outcomes include:
- Child-centred focus: Prioritising the child's well-being by considering all caregivers in their network, not just family.
- Strengthening existing support: Identifying and leveraging positive connections already established in the child's life.
- Empowering social workers: Providing information and insights to support their decisions, not replace their expertise.
- Improved access to information: Guiding social workers to find relevant information quickly and efficiently.
- Ethical and professional use: Emphasising responsible application aligned with social work ethics and best practices.
- Co-creation with social workers: Designing the tool with social worker input to ensure it assists rather than hinders their work.
- Data-driven, person-centred approach: Combining data analysis with social workers' critical thinking for informed decisions.
Proof-of-concept benefits
North Yorkshire Council's proof-of-concept project has yielded significant benefits for both children's social care and social workers themselves. By leveraging Azure AI technology, the project developed a tool that can:
- Analyse information from various sources (notes, images, system data) within the existing case management system. This empowers social workers to identify connections between cases and individuals that might be missed otherwise.
- The pilot demonstrated a 90 per cent reduction in time and cost for some data retrieval tasks compared to the current system. It also quantified the previously unmeasured burden of data retrieval on social workers.
- Social workers have the ability to automatically generate eco-maps that visualise the networks surrounding children and families. These can reveal hidden connections that could take days to discover manually, enabling proactive safeguarding and better network involvement.
- Historical data remains accessible and understandable, even when the original staff member is no longer works for the council or unavailable.
The project's benefits extend beyond improved efficiency. By minimising administrative burdens, social workers can dedicate more time to direct engagement with families, building stronger relationships and potentially leading to better outcomes for children and families. The tool facilitates early risk identification, allowing for proactive interventions that improve child safety and well-being. Furthermore, user-friendly tools address frustrations with the current system, boosting worker satisfaction. Positive feedback from social workers and stakeholders highlights the project's potential to address key challenges in children's social care.
The project, funded by the Department for Education's Data and Digital Solutions Fund (ended March 31, 2024), employed a robust benefits mapping and realization approach to ensure efficient resource allocation. Future plans can leverage these successful outcomes to secure further support and wider implementation.
Technical hurdles
North Yorkshire's proof of concept provided valuable learnings as it navigated some technical hurdles. These challenges impacted the initial scope of user testing and implementation timelines. While the tool wasn't as feature-rich as initially envisioned, the council prioritised core functionalities, establishing a solid foundation for future development. Data refresh, data matching, LiquidLogic form integration, and auto-redaction are exciting possibilities the council will explore in subsequent phases.
Throughout the project, the Board and key stakeholders were kept informed, and these adjustments were deemed necessary for successful exploration within a new and innovative area.
Next steps
North Yorkshire Council's successful proof-of-concept has positioned it at the forefront of AI-powered solutions for children's social care. While further development is necessary, the project's advancements surpass existing workarounds within the sector.
Overview
In November 2023, Somerset Council took a pioneering step by launching a discovery project to explore Microsoft Copilot. The pilot aimed to gain early insights into the potential of AI to transform the council, focusing on improving workforce efficiency, aligning with overall council goals.
The council established a lean project team and adopted an organic approach to identifying use cases. Somerset offered 300 Copilot licenses to volunteers across the organisation eager to trial the technology in their day-to-day work. They quickly distributed all 300 licences, allowing them to gather real-world data and explore safe implementation strategies for AI within the organisation.
Somerset Council are part of the Microsoft Copilot Early Access Programme (EAP) alongside 11 other councils. This innovative programme grants them access to cutting-edge Copilot updates and dedicated Microsoft support.
AI governance
Ensuring AI's safe and ethical use was paramount to the council’s discovery project. Somerset placed data protection, ethics, acceptable use policies, controls, and user education at the heart of their rigorous AI governance strategy. They developed data protection impact assessments (DPIAs), ethical guidance, and a clear acceptable use policy.
The council developed their governance framework alongside the pilot, incorporating valuable feedback from volunteers. This iterative approach ensured the framework addressed real-world use cases and remained relevant and up to date.
The core project team acts as the first point of contact for escalations and is actively developing governance and compliance policies. This team collaborates with the Integrated Care Board (ICB) and Digital Transformation Board, keeping them informed and ensuring alignment on AI principles across the council and with health partners. An additional sub-group specifically focuses on AI ethics and information governance.
The council actively strengthens its governance approach through the local government collaboration and innovation copilot subgroup, a monthly council-led forum that Somerset co-leads. Convening over 300 knowledgeable council representatives, this forum fosters a collaborative environment for knowledge sharing on copilot. Through open discussions, councils leverage each other's experiences to tackle challenges collectively and develop unified governance practices.
Looking ahead, the council plans to establish a cross organisation AI board to provide strategic direction beyond the discovery project. This board will comprise senior management, service area representatives, information governance experts, technical architects, and external partners. Additionally, Somerset is developing comprehensive staff training to promote the responsible use of AI.
Support and training
Somerset Council provided comprehensive support to staff throughout the AI discovery project, fostering a collaborative learning environment. Participants received ongoing support through a self-sustaining "pioneer community" on Microsoft Teams. This tech-savvy group, facilitated by the project team, shared questions, answers, and experiences, fostering a spirit of experimentation and support.
Complementing the pioneer community, the project team hosted weekly contact meetings. Sessions initially focused on updates, demonstrations, and addressing user feedback. As the group's expertise have grown, the meetings have evolved into more community-driven discussions with specific themes. The change in these meetings, reflects the council’s advancement with Copilot utilisation.
Looking ahead, Somerset plans to establish dedicated forums tailored to departmental or user group needs. These structured forums will cater to specific goals and facilitate focused discussion on specific applications of Copilot. Additionally, the council is exploring the potential of a future "Copilot Champion" scheme to further empower and support staff.
Early benefits and return on investment
Somerset Council's commitment to monitoring the value of Copilot has yielded positive results. From the outset, they have continually engaged with staff using Copilot, gathering detailed feedback through surveys, interviews and usage data. The council established a system for measuring benefits through regular three-month reviews, capturing both qualitative and quantitative data.
Their February checkpoint review, three months into the discovery, highlighted significant improvements in staff productivity:
- Copilot significantly reduced time spent drafting meeting minutes and actions. Staff reported a time saving of 55 minutes, going from 72 minutes to just 17 minutes with Copilot's assistance.
- Notably, 88 per cent of neurodiverse users experienced productivity gains, particularly during meetings and document creation.
- Overall, 75 per cent of users reported improved focus, and 53 per cent found their work more enjoyable with Copilot.
On average, Copilot users saved between 2 and 4 hours per week. By calculating this time saving against the monthly Copilot cost (£25), Somerset has achieved a return on investment (ROI) of 4 to 10 times. These initial benefits stemmed from simple uses like meeting recording and note generation, without formal training. Identifying these natural use cases quickly demonstrated a positive ROI, boosting senior leadership's confidence in Copilot's further development.
This strategic approach to monitoring progress and productivity gains has provided a strong foundation for the council to explore service-specific use cases of Copilot.
Challenges and limitations
While a small number of the initial 300 Copilot licenses were returned due to unmet expectations, this is offset by a waiting list exceeding 70 staff, indicating strong overall enthusiasm for AI adoption within the council. However, monitoring usage remains a challenge. Current Microsoft dashboards are acknowledged to be inaccurate, requiring Somerset to develop several workarounds such as user surveys, adding to the project team's workload.
Despite these hurdles, Somerset leverages its early adopter status by fostering a critical partnership with Microsoft. They've identified and communicated multiple functionality gaps between real-world use cases and advertised capabilities. This valuable feedback will be instrumental in shaping Copilot's future development, not only for Somerset but for all councils looking to adopt Copilot in the future.
Advancing Somerset’s AI journey
Leveraging their growing expertise in Copilot's capabilities, Somerset Council plans to explore how the technology can benefit specific service areas and different staff groups. This targeted approach will inform a strategic rollout, expanding Copilot licenses across the council for maximum impact.
Somerset Council's initial explorations identified specific staff groups that could benefit from AI tools, such as neurodiverse users. Additionally, the council has assessed high-pressure areas like social care and identified tasks that could be automated.
Somerset recognises that Microsoft 365 Copilot is a valuable springboard, but for complex tasks, they envisage leveraging a wider AI toolkit. This could include Azure AI and Copilot Studio as they advance on their AI journey. To maximise the value of their licences, the council intends to establish tiered access structures tailored to specific user or departmental needs. They're particularly excited about the potential for AI to unlock automation and generate advanced data insights that can transform the organisation.
Overview
Despite its size, Surrey Heath Borough Council, a non-metropolitan district council in southeast England, is a champion of digital innovation. Having successfully implemented several digital technologies in the past, the council is now on an exploratory journey to integrate Artificial Intelligence (AI) into its operations. Recognising the potential of AI to improve service delivery and streamline operations, the council is committed to understanding and leveraging this technology, even in the face of budgetary constraints.
This case study was compiled based on a discussion with senior staff members at Surrey Heath: James Rutter (ITC Manager), Sally Kipping (Head of HR, Performance and Communication) and Stuart Field (ITC Manager). The case study details the council’s early exploration of AI, specifically Microsoft Copilot.
For a small council like Surrey Heath, it is essential that every technology investment demands a clear return on investment. The council’s leadership recognises the potential of AI to transform operations, but they must balance financial responsibility and identify practical applications within the organisation.
"We're committed to continuous improvement in our use of technology," says Sally Kipping, Head of HR, Performance and Communications. "However, building a strong business case for AI is challenging, especially with our budget constraints. We need to pinpoint areas where AI can demonstrably add value, whether by enhancing resident services or tackling financial hurdles – ideally achieving both."
Echoing this sentiment, James Rutter, acknowledges the importance of AI but underscores the need for caution. "AI is still in its early stages," he explains. "While we can't afford to ignore it, we are very aware that residents, staff and users may need reassurance about the use of AI and taking unsubstantiated risks isn't an option. We need a measured approach to explore its potential benefits."
Exploring AI
To begin their AI exploration, Surrey Heath Council has strategically targeted a few applications. Leveraging two Microsoft Copilot licences issued to senior IT staff, the council has initiated a low-risk experiment to assess AI's potential value within the organisation. This initial phase focuses on testing Copilot's capabilities in specific areas: summarising emails, generating draft policy documents based on prompts and notes, and creating presentations from short instructions. Through this controlled exploration, the council hopes to pinpoint where AI can bring the most significant benefits.
Initial tests with Copilot have been promising, particularly the intelligent meeting recap feature and the ability to summarise lengthy email threads. These features demonstrate the early value of AI integration, even before a full-scale rollout.
While Surrey Heath is excited about the potential of AI, they're taking a thoughtful approach to Copilot adoption. The council wants to ensure a smooth and successful integration, so they're carefully considering factors like training needs, organisational readiness and return on investment.
"Innovation is one of our core values," James Rutter explains regarding their measured approach with Copilot licenses. "However, we prioritise strategic implementation. Unlike larger councils, we have to be mindful of resource allocation. By starting with a limited group of users, particularly within management, we can closely monitor the impact and build a strong foundation for wider adoption in the future."
Continuing Surrey Heath’s AI journey
The council has identified several promising areas where Copilot can enhance productivity based on initial testing. They plan to gradually expand the number of Copilot licenses as part of their ongoing exploration. Licenses will be distributed strategically to the wider management team, with continued monitoring to ensure a successful and measurable rollout across the council.
While Copilot offers a range of exciting possibilities, the council acknowledges its limitations. As James Rutter explains, "One example might be to use AI for minute-taking. However, whilst AI can accurately transcribe meetings, it currently struggles with context and nuance, which are essential for capturing meaningful minutes."
Surrey Heath Council is clear about Copilot's role as a tool working with staff to empower and support them. The goal is to leverage Copilot's capabilities to enhance productivity and reduce administration to free up valuable time for employees to focus on more meaningful and strategic tasks to increase services to residents and increase role satisfaction.
Surrey Heath’s vision extends beyond Copilot. They're exploring a strategic migration of their Content Management System (CMS) to SharePoint and OneDrive. This move would not only unlock deeper integration with Microsoft products like Copilot, but also generate cost savings by streamlining existing subscriptions.
Looking ahead, Surrey Heath Council is actively seeking to learn from the expertise of the Local Government Association (LGA). They're particularly interested in understanding how other councils are approaching AI governance, exploring different implementation strategies, and identifying areas with the highest impact. The LGA offers valuable resources, including the regular AI Network, which convenes every six weeks to delve into various AI themes. This forum allows councils to consider diverse aspects of AI adoption and deployment.
To further enhance knowledge sharing, the LGA has recently launched a dedicated Microsoft Teams channel for councils to exchange key documents and ideas. Additionally, we have established a new AI practitioners’ network specifically for advanced AI users working on innovative AI applications.
Paving the way for district councils
Surrey Heath Council's exploration of AI showcases the diverse approaches within the local government sector. This case study exemplifies how smaller councils can strategically experiment with AI without straining resources. By collaborating with peers and learning from fellow councils, they aim to integrate AI for operational efficiency and resident benefit.
Surrey Heath's journey provides valuable insights for other smaller councils considering AI adoption. Their approach underscores the importance of thoughtful planning, financial prudence, and collaborative learning.
Overview
Swindon Borough Council boasts a forward-thinking approach to technology. Their dedicated Emerging Technology team has leveraged AI for several years, delivering cutting-edge solutions that address service-area challenges. The council first considered AI over five years ago and have continued to trial and innovate with AI, recently deploying a solution to cut their translation services by 99.6 per cent.
Identifying another crucial area, the council focused on creating accessible information for residents with learning disabilities. The Accessible Information Standard (AIS), introduced in 2016, mandates a consistent approach for all relevant organisations. Additionally, the Equalities Act 2010 emphasises easy-read documentation as a reasonable adjustment for people with learning disabilities.
However, existing methods for generating easy-read documents were expensive and slow. Each A4 page cost around £120 and documents took weeks to complete, or social workers had to dedicate significant time to create them manually. This hindered the council's goal of empowering people with learning disabilities by providing them with equal access to information and opportunities.
There are over 1 million people with learning disabilities in the UK alone. Easy-read documents, with their clear visuals, larger fonts, and concise language, ensure they can make informed decisions about their lives.
Swindon's AI solution for easy-read documents
Swindon Borough Council faced a challenge: creating accessible "Easy-Read" documents for residents with learning disabilities was time-consuming and expensive. Traditional methods for creating "Easy-Read" documents were not only time-consuming, but also financially unsustainable. Their groundbreaking solution, "Simply Readable," leverages artificial intelligence (AI) to generate these documents in a matter of minutes, at a cost of c.1p per page. This dramatic reduction in time and expense frees up valuable time for social workers, allowing them to focus on providing direct support to residents. More importantly, it ensures residents with learning disabilities have timely access to crucial information, empowering them to make informed decisions about their lives.
"Simply Readable" embodies Swindon's dedication to fostering social inclusion. Recognising the importance of resident experience, the council partnered with a group of residents with learning disabilities throughout the development process. This co-creation approach ensured the tool addressed genuine needs. The residents' expertise directly shaped key features like the use of clear, concise language, larger fonts for improved readability, increased spacing to avoid overwhelming the reader, and the inclusion of relevant images to enhance understanding. This ongoing collaboration with the resident group serves as a feedback loop, guaranteeing "Simply Readable" remains effective and meets the evolving needs of its users.
The user-friendly Amazon Bedrock platform provided the foundation for "Simply Readable." This platform's intuitive interface allowed the council access of choice to pre-built foundation models, to develop and customise their preferred AI model without requiring extensive coding expertise. Social workers can now simply input complex information and receive easy-to-understand versions in under 30 minutes, significantly streamlining communication. The impact of "Simply Readable" extends beyond residents with learning disabilities. By integrating with existing translation services, the tool can translate documents into a staggering 75 languages. This expanded functionality benefits non-native English speakers and individuals with lower literacy levels, promoting greater inclusivity within the community.
Governance
They boast several years of experience with AI, leading to a well-developed governance framework, including a specific AI governance board and a set of AI Principles based on Asilomar. The council also engages their information governance team, adhering to data protection regulations through self-written Data Protection Impact Assessments (DPIAs).
Uniquely, they prioritise identifying and mitigating potential unintended consequences of AI deployments. This involves a proactive "unintended consequence scanning" process conducted both internally by their team and collaboratively with colleagues from other departments venturing into new projects (like Adult Social Care). It's important to note that this goes beyond just AI; it's a broader risk management approach applicable to any initiative; they also undertake Equality Impact Assessments for all work.
While they have a dedicated AI ethics board/group, the unintended consequence scanning extends beyond its purview, demonstrating a layered governance approach. This suggests the organisation has a formal structure for overseeing AI development and deployment, with the ethics board likely focusing on higher-level principles.
Simply Readable benefits:
- Simply Readable significantly reduces the financial burden of creating Easy-Read documents. Previously, a complex 5-10 page document could cost upwards of £600 to convert. With Simply Readable, costs have reduced to just 7-10 pence, delivering significant savings that can be redirected towards other critical social services.
- The tool streamlines the easy-read document creation process, freeing up valuable time for social workers. Now, easy-read documents can be created in under 30 minutes, allowing them to dedicate more time to providing direct support, to residents within the community.
- Simply Readable is a cloud-based solution, offering a user-friendly interface that requires minimal technical expertise or ongoing maintenance. Furthermore, the absence of per-user licensing fees makes it a cost-effective and accessible option for wider adoption.
- The development process for "Simply Readable" heavily emphasised user feedback and co-creation with residents with learning disabilities. This ensures the tool directly addresses their needs and preferences.
- Swindon Borough Council's commitment to social good extends beyond their local community. Their decision to make "Simply Readable" open-source allows fellow councils to leverage this innovative tool.
Measuring return on investment (ROI):
Swindon Borough Council implemented a data-driven approach to comprehensively assess the effectiveness and ROI of "Simply Readable." This involved a meticulous three-month baseline period where they established key metrics for factors like document creation time, personnel costs associated with manual conversions, and resident feedback on accessibility and comprehension. By meticulously tracking both quantifiable benefits (e.g., reduced costs, staff time saved, improved resident satisfaction) and user experience throughout the project, the council gained valuable insights to inform future initiatives.
The impressive 749,900.00 per cent figure for Adult Social care takes into account the initial investment in developing the tool, the ongoing operational costs (cloud hosting, maintenance), and the quantifiable benefits outlined above. This staggering figure underscores the transformative potential of "Simply Readable" and AI-powered solutions in promoting social inclusion while achieving significant cost savings and improved operational efficiency. Their Translation tool, also open-sourced has an ROI of 6,300,000 per cent.
What next?
Following successful deployment of AI, Swindon are currently exploring the technical achievability and business benefits in using a multi-AI based solution to record, transcribe and document conversations and associated actions discussed between Adult Social Care professionals and those they support, such as Care Act and Mental Capacity Assessment conversations.
Overview
Warwickshire Council has been a pioneer in applying artificial intelligence (AI) within local government. Their journey began in 2018 with a project utilising convolutional neural networks to analyse video data for traffic management. This initiative, alongside subsequent projects like email clustering using machine learning, demonstrated the value of AI for gleaning insights and improving service delivery.
Warwickshire's latest project, "Access to Food," tackles a pressing issue in the face of the rising cost of living. It aims to understand citizens' access to healthy food options, particularly in deprived areas where supermarkets might not be accessible. The project goes beyond simply identifying shops; it seeks to classify businesses based on factors such as size and product availability. This approach reveals potential areas with limited access to affordable and nutritious food.
While traditional algorithmic methods could achieve this task, Warwickshire opted to explore the potential of Large Language Models (LLMs). Instead of relying on pre-defined rules, they wanted to assess LLMs' ability to learn and classify businesses based on a general overview. This innovative approach involved utilising the council's in-house expertise to download and implement the open-source LLM, Lama 2. This ensured complete data sovereignty and minimised risks associated with third-party data handling, making the proof-of-concept a low-risk venture.
The project prioritised developing a robust data infrastructure to optimise LLM performance. This emphasis on in-house expertise and data security reflects Warwickshire's commitment to responsible AI adoption while exploring its potential for citizen services.
Governance and internal engagement
To ensure responsible AI development, in-house projects undergo a dedicated review process. This includes checking against a Data Protection Impact Assessment (DPIA), an Equality Impact Assessment (EIA), and a data ethics framework, depending on the project's nature. Additionally, a data readiness scoping exercise identifies potential data quality issues, ensuring the project uses viable data.
This project involved collaboration with wider Business Intelligence (BI) teams. An evaluation report summarising the findings was shared with the BI service in strategic research and further discussed with other service areas for potential integration.
A usage policy for Enterprise CoPilot provides guidance on responsible AI use. Data ethics and algorithmic transparency are key priorities on the council's data roadmap for this year. To further strengthen governance, oversight boards are planned for implementation by year-end. Additionally, the Head of Data will assume the role of AI Lead, overseeing AI governance and implementation across the council. Finally, the council plans to engage its Citizens Panel on the use of AI within local government.
Proof of concept: promising results but room for improvement
While the initial AI implementation showed promise, a simpler data science approach proved more accurate for this specific task. The AI model achieved 63 per cent accuracy based solely on prompt engineering, a more straightforward approach using string concatenation (without AI) achieved 95 per cent. However, the AI performed well at understanding the general task, with 96 per cent accuracy, but struggled with the specific classification challenge. Despite limitations in accuracy, the model shows promise for saving staff time by automating multiple manual searches through further fine-tuning.
Challenges and learnings
Several challenges arose from the council's approach to develop and manage their own LLM:
- The council lacked a dedicated funding stream for AI solutions. As a workaround, they used the most powerful local laptop at their disposal to run the LLM.
- The chosen LLM (Llama2-12b) required control through human reinforcement and efficient performance with data compression (quantisation). They encountered challenges configuring it for their hardware, leading to slow processing times (up to 1.5 hours per output). A separate cost analysis revealed the significant expense of dedicated LLM hardware, with a used graphics card costing around £41,000.
- Security concerns arose during the project. Despite instructions to only provide classifications, the AI generated irrelevant information (hallucinations) in 3.8 per cent of cases. Eight instances presented a potential security risk by including business names in the responses. This highlights the need for robust security measures when deploying LLMs. Since this was a proof-of-concept with internal data management, risks were mitigated within the council.
Data quality was satisfactory for this project, focusing solely on business properties and types. Despite the challenges, the development cost was minimal, requiring mostly time investment. Overall, the project demonstrated the potential of LLMs but also highlighted the need for careful consideration of accuracy, security, and hardware requirements before full-scale deployment.
Overall and what next
Warwickshire Council's "Access to Food" project serves as a bold experiment in exploring open-source Large Language Models (LLMs) for local government applications. While simpler methods achieved superior accuracy for the specific task of business classification, the project yielded valuable learnings.
The council's innovative decision to leverage an open-source LLM (Lama 2) demonstrated a cost-effective and data-sovereign approach. This strategy minimised risks associated with third-party data handling and empowered the council to explore the technology independently. However, the project also highlighted the challenges associated with open-source LLMs. Configuring Lama 2 for their hardware proved complex, and processing times were slow.
Overall, Warwickshire Council's approach provides valuable insights for local governments considering LLMs. Their exploration in leveraging open-source models, and identifying limitations like security risks and hardware demands, paves the way for further responsible exploration of open-source LLMs in local government services. The council continues to explore cost-effective ways to run LLMs in-house for future applications.
West Berkshire Council: Development of generative AI tools
West Berkshire Council are experimenting with the use of generative AI within services. Leveraging existing skills within the council, they are taking a pioneering approach by developing their own custom AI tools in-house built on commoditised large language models. This strategy allows them to explore cost-effective and highly adaptable solutions that directly address their specific needs. By building their own AI tools using the ChatGPT API, West Berkshire Council retains greater control and flexibility, ensuring the AI integrates with their existing workflows and data infrastructure.
Overview
The council initially focused on streamlining the creation of job adverts and LinkedIn posts to expedite the recruitment process. They developed a prototype using OpenAI's GPT models, enabling the automation of content generation. Based on survey feedback senior managers were impressed, rating the AI-generated content at an average of 8 out of 10. This new system integrated with the council's existing digital platform, allowing for the automated gathering of information required for job adverts. Through forms, details like job titles, salaries, locations, and reporting lines were collected and then fed into the OpenAI API to generate the advert itself.
West Berkshire Council delved further, exploring "assistants," a feature offered by GPT-3.5. Assistants can be customised by feeding them relevant documents and prompts. Different chatbots for specific purposes, such as answering frequently asked recycling questions are currently being tested. Additionally, internal policy assistant chatbots are under development. This tool will enhance access to policy information, allowing employees to find swift and accurate answers to their policy-related queries.
Overcoming challenges and ensuring ethical implementation
Initially, the council encountered challenges using ChatGPT directly, finding the results unreliable. Security concerns regarding data privacy and OpenAI's data usage practices also arose. Furthermore, the recycling information chatbot, due to its limited knowledge base, provided inaccurate information at times.
To address these issues, the council strategically shifted to using secure OpenAI APIs, offering improved reliability and enhanced data protection. As part of the pilot, public data, such as existing job adverts, was utilised to minimise privacy risks. Furthermore, the AI models were meticulously fine-tuned with relevant training data (for instance, model job adverts) to bolster accuracy and guarantee consistent content that included all essential details. A data privacy impact assessment was conducted, and legal checks ensured complete compliance with all relevant regulations.
West Berkshire Council prioritised trust and ethical implementation from the outset. Recruiting managers from across the organisation actively participated in testing and refining the AI-generated job adverts, ensuring they met the council's specific requirements. The project explored applications beyond recruitment, with a focus on content simplification tools and internal policy assistants. Ultimately, the aim is to significantly improve internal communication and accessibility for all staff members.
The council has plans to establish a dedicated AI Governance Board which will comprise representatives from crucial departments like digital, ICT, legal, and procurement, as well as front-line services, and may also include a council member. Their primary function will be to ensure comprehensive oversight and the use of AI technologies within the council.
Impact and a promising future
The implementation of Gen AI has had a significant impact. Managers reported a substantial reduction in the time it takes to draft job postings. The use of AI has also ensured consistent formatting and the inclusion of all essential information, leading to a more uniform presentation of job adverts across the council.
West Berkshire Council has proven that developing custom AI tools using the ChatGPT API is highly cost-effective compared to off-the-shelf solutions. The initial projects cost a mere £300 in staff time to set up, and around £2 a month to run, demonstrating the affordability of this approach. This method also offers greater flexibility and avoids vendor lock-in with ongoing subscription fees. The project has highlighted the potential of AI to simplify complex content, automate responses to policy-related queries, and offer support for various communication needs.
Looking ahead, West Berkshire Council is committed to expanding its Gen AI adoption. They plan to develop new AI tools and integrate them into a wider range of processes. Additionally, they are exploring off-the-shelf AI products such as Microsoft's Copilot and Amazon Bedrock to assess their suitability for specific needs. This multi-faceted approach fosters flexibility, grants the council control over their data, and ensures they identify the most effective AI solutions for their diverse council requirements. Finally, the council intends to expand the number of internal policy assistant chatbot to encompass a broader range of policies, offering employees even more comprehensive support in navigating policy-related information.
West Berkshire Council's innovative approach to Gen AI adoption serves as a model for local government bodies seeking to optimise workflows, enhance communication, and embrace technological advancements for the benefit of their employees and the communities they serve.
If your council is adopting or implementing AI and would like to be featured on our use case bank, please contact Tom Hindmarch.