The Local Government Association (LGA) has said councils using agentic AI should be ambitious but risk aware in new guidance on adopting the technology.
The guidance, A playbook for agentic AI in local government, includes an explainer on the technology, details the known risks associated with it, and recommends a series of adoption principles.
Agentic AI is an emerging technology that can be given a goal, break the goal in to steps, autonomously use approved tools or data sources, check progress and carry out actions on behalf of a person or organisation.
According to the document, local government is increasingly considering the technology as a tool to boost efficiency, automate routine tasks and improve service delivery.
But implementation comes with risks that must be understood and managed from the outset, the guidance warns.
The document lists nine potential risks that councils using the technology may face, including a warning that AI agents could be vulnerable to malicious prompts, such as instructions to disregard previous safety protocols or to gather and share sensitive client data. Such actions can lead to unethical behaviour and data breaches, it says.
It also warns that constant human oversight is not always feasible as AI agents, meaning that a degree of trust is required.
The technology also carries risks relating to accountability, errors and bias, and costs, among others.
To tackles these risks, the guidance sets out 11 recommended principles that councils adopting agentic AI should follow, including on procurement, accountability, training, cyber security controls.
The report says: "Agentic AI has the potential to transform local government operations, but only if it is adopted deliberately, transparently and with people remaining clearly accountable.
"Councils should be ambitious in exploring the opportunities, while also being disciplined about risk, data, ethics, workforce impact, cyber security and public trust.”
It adds: "Agentic AI adoption will depend as much on data maturity, process clarity and leadership discipline as on technical capability. Councils that invest in clear ownership, quality data, evidence-based decisions and accountable governance will be better placed to use agents safely, effectively and in a way that strengthens public trust.
"Agentic AI is not primarily a technology challenge. It is a leadership, governance, workforce and service design challenge. Councils which invest in data quality, accountability and responsible governance will be best placed to realise the benefits, protect public trust and create sustainable public value."
