Generative AI has a wide potential for local authorities, including responding to natural language questions, understanding different types of data and producing outputs such as text and code, according to the London Office of Technology and Innovation (LOTI).
It has published a collection of guidance documents on the use of the technology, developed with AI software designer Faculty AI and aimed at local authority leaders, chief information officers and staff, along with a ‘state of play’ in local government.
Along with highlighting how generative AI – given the term genAI – could be used, they emphasise the need for ethical considerations, data privacy, transparency and trust.
The document for CIOs – the most detailed of the collection – says councils should focus on outcomes rather than using genAI for its own sake, and keep in mind it is developing quickly so they will need to review use cases and opportunities frequently. They will also have to watch for relevant legislation on how it can be used.
It outlines use cases in research and writing, as a personal assistant, in generating images and marketing ideas, supporting data analysis and providing coding support. The latter would involve helping detailed processes such as debugging sections of code and converting code from one programming language to another.
Need for caution
But these come with a note of caution, as LOTI suggests local authorities approach high risk use cases carefully. This could involve the use of resources such as safe testing environments and synthetic datasets to train AI.
The document for leaders points to the potential for applying genAI to internal data for purposes such as summarising long documents, call centre logs and case notes, analysing language across documents and automating the drafting of low risk letters.
As a personal task assistant it could be used in ideation, drafting, structuring new documents, checking gaps in thinking, creating images for communications teams and assisting coding.
But harnessing the potential requires efforts in building skills among staff, preparing data, developing governance and rethinking service delivery, LOTI says. These should be accompanied by ensuring the use of generative AI is safe and fair, transparent, accountable, and private and secure.
The ethical issues are reiterated in the document for council officers, along with tips on writing good questions for AI. These including giving specific instructions, breaking down tasks into stages and sharing an outline of what you want it to say before asking for outputs.
According to the state of play document, a survey with 37 respondents indicated it is still early days for the deployment of generative AI in local government, with more than a third not yet aware of it being used in their organisations.
But there are signs of a few councils beginning to use it in writing and editing documents, and to generate and check code to improve the quality of software applications.
LOTI – a membership organisation for London boroughs – said the guidance has been created from interviews with council officers and experts on AI, along with roundtables and desk research.