Agency is using Microsoft Azure-powered “Investigations Assistant” to check spending decisions made on behalf of people covered by a lasting power of attorney because they cannot look after their own finances
The Office of the Public Guardian is trialling the use of an artificial intelligence tool to speed up investigations into concerns about financial misappropriation or abuse committed against people who are the subject of a lasting power of attorney.
The OPG, which is an executive agency sponsored by the Ministry of Justice, exists to look out for the interests of people in England and Wales who cannot make important decisions for themselves. When such individuals are the subject of a lasting power of attorney – or an enduring power of attorney – and concerns are raised about the potential misuse of those arrangements, OPG has the power to investigate.
According to the OPG’s annual report and accounts for 2025-26, a total of 3,645 concerns raised with the agency were accepted into investigations over the course of the year.
A transparency document published this month by the Cabinet Office and the Department for Science, Innovation and Technology says the OPG has developed a digital tool to support investigators analysing and categorising financial data from bank statements provided to the agency in response to concerns being raised.
It says the “Investigations Assistant” is an AI‑enabled workflow designed to support checking processes when concerns raised relating to financial misappropriation or abuse trigger an investigation.
Related content
Student Loans Company sheds light on automated decision-making for finance applicationsHMPPS sheds light on prisoner ‘violence estimator’ algorithmHMRC algorithm for automated council tax valuations offers ‘greater consistency and quality’
The automated assistant supports “routine, time‑consuming tasks” that investigations and support staff currently carry out manually, like categorising financial transactions detailed on bank statements, such as payments for utility bills, care fees and cash withdrawals. Identifying spending that may not be in the best interests of an individual subject to an LPA or EPA is the purpose of the categorisation.
OPG’s entry for the project on the government’s Algorithmic Transparency Recording Standard Hub says that initial testing and evaluation for the tool found it “almost halved” the time it would take investigators to manually categorise transactions.
It adds that the time savings “will now be further validated in a controlled live pilot”. In addition to reducing manual effort, other benefits of the tool are said to include improved consistency and clearer audit trails.
The OPG says Microsoft supported the development of the tool by providing technical expertise and configuring the underlying AI services it uses. The agency said this included setting up secure Azure‑based components, helping to build the AI workflows such as document processing, financial transaction categorisation and ensuring the system is deployed safely. OPG said Microsoft has no access to the data used by the tool.
In addition to categorising financial transactions, the tool highlights individual items that might require further investigation.
Investigators review all AI outputs before anything is progressed or shared. They also check the categorisation of transactions for accuracy, correct any errors and complete the categorisation for any flagged transactions.
According to the OPG, “human review is the formal assurance step”. It says “nothing produced by the tool is used without an investigator’s oversight and sign‑off”.
The transparency document does not give a timescale for full adoption of the AI tool, but confirms that it is “designed for day‑to‑day use across investigations” and expected to be used to support investigations that involve financial categorisation and later analysis.
The OPG says staff training will be rolled out and that use of the tool will “scale gradually through the pilot phase and expand as staff become familiar with the workflow and ongoing evaluation of the tool/workflows”.
PublicTechnology offered the OPG the opportunity to comment further on the development of the AI tool.
It stressed that AI output produced by the assistant is treated as an initial categorisation and that investigators are required to verify, amend, or reject categorisations – and retain responsibility for identifying transactions that warrant further enquiries.
An OPG spokesperson said: “We’re committed to tackling financial mismanagement or abuse under lasting powers of attorney. This AI tool supports staff to improve analysis, speed up investigations and ensure better protection for vulnerable people. All insights are reviewed by a human investigator who is responsible for decision-making.”