Agenda item

This report provides a brief overview of the current state and future implications of generative AI technologies.

The Chair outlined the theme of the discussion and noted that the item was intended to support the early exploration of Generative Artificial Intelligence (AI) and its uses.

The Chair welcomed the following guests to the meeting:

• Jeremy Tolmie (Microsoft)
• Stelios Zarras (Microsoft)
• Darius Toomer (Agilisys)
• Eugene O'Driscoll (Agilisys)
• Professor Aldo Faisal (Professor of AI & Neuroscience at Imperial)

Tina Akpogheneta (Chief Digital Officer) addressed the meeting and noted that the Council had been working on the issue of generative AI since last summer. Digital Services had set up a working group to look at the implications and how to safely leverage the technology. The Council wanted to explore carefully and had taken a measured approach. For ‘tech week’ (10-14 June 2024) speakers from Microsoft and Gartner had been invited to discuss potential use cases. Digital Services had recently unlocked starter licences to Microsoft Copilot (a generative AI chatbot) for staff and was gathering feedback from technology champions across departments. Microsoft had also established a forum for local authorities to help them learn together and share knowledge and best practices.

Jeremy Tolmie and Stelios Zarras from Microsoft gave a presentation covering the following:

• An overview of the history of machine learning, deep learning, and generative AI.
• The key uses of generative AI – content generation, summarisation, code generation, and semantic search.
• Applying generative AI to business problems to increase productivity, automate processes, improve the customer experience, and build creative content.
• Responsible AI principles – privacy and security, inclusiveness, accountability, transparency, fairness, and reliability.

Darius Toomer and Eugene O'Driscoll from Agilisys addressed the meeting and covered the following points:

• Agilysis had been working with the Council to identify use cases such as personal productivity and handling complex processes and applications.
• The importance of a learning and development culture that requires new teaching and learning.
• A culture of innovation and continuous improvement, which Agilisys provides to local councils.
• The need for strong data governance and infrastructure as an enabler for AI services.
• AI being used to search and make use of unstructured data – which offered the public sector an array of new opportunities, for example using predictive analytics in social care services.

Professor Aldo Faisal (Professor of AI & Neuroscience at Imperial) said AI technologies represented a revolution that would transform white collar work. In local government he saw a number of opportunities given the large datasets available and various types of decisions that lent themselves to AI. Regarding risks and opportunities he noted the risk of manipulation, but also the benefits of better, more granular decision-making and messaging.

Councillor Rowbottom noted she was supportive of the agenda but had the following questions:

• What was the timeline for implementation of the Council’s AI working group?
• Regarding use cases, what was the approach to complex cases in Adult Social Care and Children’s Services where the majority of spend goes?
• What were the Council’s thoughts on improving the interoperability of data through genuine co-commissioning with the NHS other local services?

Tina Akpogheneta said the Digital Services working group had been trialling Microsoft Copilot and other generative AI technologies and were in the process of familiarising staff with these tools, through a package of support. Opportunities were being explored in the Housing department, in contact centres, and in the Resident Experience and Access Programme (REAP). Requests for these tools used the standard H&F business case process.

Councillor Helen Rowbottom said she was interested in the use of these tools in complex cases but warned that any use of these tools for resident facing work had to be carefully evaluated and monitored.

Councillor Rowbottom asked how areas of priority were fed through to the programme. Tina Akpogheneta said the process was to put forward a business case then come to the working group for approval or amendments.

Darius Toomer said there were use cases for social care, though they required more governance – e.g. safeguarding meeting minutes. He said there was a lot of business administration work that could be automated, including lower risk social care documentation. He added that one of the main opportunities was a new era of generative AI chatbots that could be used to triage resident queries to the relevant teams.

Stelios Zarras said people should think of AI as a technology layer with a wide range of uses. He felt just looking at return on investment was too limited a view – asking how could a value be put on saving lives or helping children. He encouraged councils to be ambitious and noted some were already moving at pace.

Professor Aldo Faisal noted that if citizens were interacting with these systems, they needed to understand how their data was being used. There had been problems reported in other countries around prejudice based on the underlying data so organisations should be mindful. They may want to know what data had been used to train the models, how systems were used by people who speak English differently, and what values the models were being aligned to.

The Chair asked how quickly businesses cases were being evaluated by the working group and what framework was being used to assess against. Tina Akpogheneta said they looked at the expected return on investment and the financials of the service, as well as wider considerations.

Councillor Nikos Souslous asked how AI fit with the Council’s commitment to digital accessibility. He also asked if there was a national AI framework to help (or hinder) Councils in this work. Tina Akpogheneta said there was legislation proposed around automated decision making, but the Council was trying to think through potential risks, ahead of any future legislation. The Council was looking at developing a set of standards.

Councillor Florian Chevoppe-Verdier (Lead Member for European Co-operation and Digital Innovation) added that the Council co-produced digital services with residents to ensure digital exclusion was minimised.

Councillor Victoria Brocklebank-Fowler asked how big the AI working group was. Tina Akpogheneta said it was comprised of eight standing members who met every fortnight.

Councillor Brocklebank-Fowler asked if the working group had the necessary expertise to innovate and take action in response to requests from across Council departments.

Tina Akpogheneta said officers had proactively set up the working group in response to this emerging technology. Officers started researching and testing products on the market. They put together a database of use cases and how to assess them. The team had guidance on how to deploy and scale up these technologies in services.

Councillor Brocklebank-Fowler asked Agilysys about their work with other councils. Darius Toomer said they worked with a range of councils. For example, in the Manchester region some council development partners were trialling generative AI in areas like social care, freedom of information requests, and complaints.

The Chair asked how many councils they worked with in total. Darius Toomer said they worked with 60 councils total, with 15 engaged generative AI councils, and four development partners.

Councillor Rory Vaughan asked about the volume of data needed for accurate and meaningful predictions. He also raised the question of biased outcomes if the underlying data contained biases.

Darius Toomer said for predictive analytics there wasn’t a specific volume of data required, but to see if projects were viable they carried out a discovery project using anonymised data to see if there were the necessary markers that could be used for predictions.

Stelios Zarras noted that the use of AI in relation to social care was not restricted to predictions. He gave the example of a field worker who could use AI to make them more efficient by typing up their notes automatically or taking dictation. He added that when considering the data required for these systems to work well, it was less about volume and more about quality.

Councillor Vaughan asked if officers had used generative AI to write council reports or minutes. Tina Akpogheneta said no, not yet.

Councillor Vaughan asked about the timelines for using this technology across the Council. Tina Akpogheneta said officers were rolling out Microsoft Copilot for Edge and conversations about other programmes were starting.

Councillor Nicole Trehy asked about the implications for staff and companies. She also raised concerns about the dangers of people trusting these systems over other humans. Stelios Zarras acknowledged the concerns about the impact on jobs but stressed that Microsoft saw the technology as a force multiplier for human workers, not a replacement. Professor Aldo Faisal felt the answer to Councillor Trehy’s second point was education. People needed to understand if these systems were producing accurate information, if they could be trusted correct, and when to challenge their outputs.

Councillor Jacolyn Daly asked about how to move forward ethically and safely. She asked if other organisations were looking at this and if there was any learning to take that the Council could put in place. Professor Aldo Faisal offered his advice and support to the Council in this area.

The Chair then invited reflections from Councillor Rowan Ree, Cabinet Member for Finance and Reform. Councillor Rowan Ree addressed the meeting and thanked the guests and committee members for attending. He said this technology was developing rapidly and all areas of society were rushing to harness the benefits of AI. He wanted to ensure H&F was embracing opportunities if they led to better, more efficient services for residents. He felt there were huge opportunities with AI and other technologies and noted the investments the Council had already made to improve efficiency including robotic process automation and business intelligence. He also highlighted the co-production work on digital accessibility.

Councillor Ree noted that the Council had a huge amount of data across different services and said he was keen to ensure that it was used in the most effective way to design services to address people’s lives as they are. He welcomed Professor Aldo Faisal offer for further discussion and support.

The Chair summarised the discussion and noted the following points:

• This was the beginning of a discussion about how the Council could harness AI technologies for the benefit of the organisation and residents.
• There were a wide range of opportunities – from simpler data processing to more complex areas like predictive modelling.
• The Council needed a structure in place to assess bids and opportunities for AI in its services.
• She was interested in looking at examples from other organisations for learning that could be applied in H&F.
• The Council needed to be mindful of the issue of bias and the quality of the underlying data powering the models.
• Any AI technologies used by the Council had to work for residents. Some education and communication may be needed to allay fears about how data was being used.

The Chair thanked Councillors, officers and guests for their contributions.

RESOLVED

1. The Policy and Oversight Board noted and commented on the report.

2. The Board requested more information on the AI working group and asked officers to think about a more formal structure going forward.

3. The Board asked officers to look at examples of use cases in other local authorities and organisations that could be applied locally.

ACTIONS: Tina Akpogheneta