The Department for Work and Pensions (DWP) has revealed it is using artificial intelligence to assist with benefit decisions through a tool called “online medical matching”.
The system helps agents assess applications for millions of people claiming benefits.
Employment Support Allowance (ESA), the main health-related benefit is claimed by 1.5 million people in the UK.
The AI tool works by comparing the health condition(s) a claimant enters when applying for the benefit with a centrally maintained list and finding the “closest match”. It reports a 87 per cent correct prediction rate for this process.
The DWP says the technology is used to determine benefit eligibility, with human agents overseeing the process.
Labour is rolling out AI across DWP services
GETTY
Since its implementation in July 2020, the AI system has processed more than 780,000 cases and saved 42,500 operational hours, according to the DWP.
However, the department emphasises that human oversight remains central to the decision-making process.
Once the AI tool matches medical conditions, the claim is registered on the ESA benefit system through automated registration.
A human agent then conducts a case review, making the final decision on whether ESA should be awarded.
The DWP stresses that agents do not simply accept the result provided by the solution.
LATEST DEVELOPMENTS:
The initial version of the tool, used from 2020 to 2024, achieved only a 35 per cent correct match rate, with agents needing to correct 65 per cent of cases.
This low success rate was attributed to the original algorithm attempting to match spellings rather than context.
Concerns about automated benefit decisions have intensified after revelations in December that a separate DWP machine-learning programme for detecting welfare fraud showed bias based on age, disability and nationality.
An internal assessment found “statistically significant outcome disparity” in the fraud detection tool’s results.
However, the DWP maintains the ESA matching tool has a low risk of bias, as it only processes medical conditions without accessing personal details.
Shelley Hopkinson, head of policy and influencing at Turn2us, said: “With many struggling to contact the DWP and facing long waits for decisions, using automation to speed up processes could be positive – if done right.
“Past issues with AI-driven decision-making show how automation can introduce bias and unfairly penalise people who need support.”
“It’s crucial that people understand how these systems work and have clear routes to challenge mistakes. The government must ensure that AI works for people, not against them.”
The DWP has responded by emphasising that the ESA matching tool processes only medical conditions, not personal details, to minimise bias risks.
To mitigate concerns about AI influence, the DWP says staff have received specific training to verify results independently
Getty
To mitigate concerns about AI influence, the DWP says staff have received specific training to verify results independently.
Agents are required to check the IRG code, which relates to health conditions and medical evidence, as part of their verification tasks.
The ESA matching tool is the first to be shared by the DWP on the government’s algorithm transparency register, despite this being a requirement for all departments.
The department maintains an internal inventory of AI tools but has resisted publishing the full list, stating that managing information distribution is “crucial for public affairs.”
