CivTech winners on empowering caseworkers with GenAI

Empathy is a critical component of casework and yet, individuals who go into this profession can spend over half their day on administration, says Angela Bradbury, COO of HelpFirst. With governance demanding consistent standards of data collection, what role can technology play in redeploying the soft skills of caseworkers away from administration and towards guiding and supporting vulnerable citizens?

Bradbury, and Andy Bell, CEO at HelpFirst, argue that in these highly stressful frontline work environments, GenAI can deliver on the radical transformation needed to create case management systems that support job satisfaction and allow citizens to gain priority and care they deserve. 

Whilst Bell has been in the tech space throughout his career, including working at a company that printed Instagram posts on fridge magnets, it was his daughter's traumatic birth that refocused his attention to the Tech4Good space. “It was the NHS at its worst,” Bell recalled.

The disorganisation of the frontline healthcare context meant his daughter was not prioritised for treatment, leading to her avoidable death.

It was after this experience that Bell recognised that without radical change to the systems and infrastructure behind frontline work, this unnecessary level of risk would remain.

GenAI on the frontline

In CivTech 8, the Scottish GovTech accelerator, Bell competed with 19 organisations to solve the challenge of how technology can be used to quickly identify and prioritise support for people in the most vulnerable situations, starting with those that had energy problems. 

The 2022 spike in the cost of living increased Ofgem’s Extra Help Unit caseload by 111 per cent on the previous year and over 400 per cent for Crisis Support, the service was falling short of delivering on service quality. Caseworkers, themselves often neurodivergent, were drowning in case notes, and risked burnout from the stress.

Tapping into GenAI’s new ability to understand the ambiguities in language, the HelpFirst solution uses AI to rigorously sift through case notes, extract and visualise risk alerts and provide a dashboard for operational delivery executives to fairly distribute cases to workers based on the severity and cases numbers in real-time.

In an earlier project, using an AI voice assistant to route calls, Bell told us, “we were able to reduce costs by 90 per cent, improve service levels and redeploy caseworkers away from admin and towards the frontline within three months”. 

Using AI in these ways is demonstrating clear benefits in alleviating the tension between the commitment to routine standards and the need for caseworkers to transfer their empathy and soft skills towards providing guidance and time to vulnerable citizens. 

Collaborative craftsmanship

One of the biggest mistakes vendors make when working with the government is falling in love with their own solution, Bradbury and Bell agreed.

“What typically happens is that vendors will do an initial scope of the requirements, make their best guess about what will match those requirements and then after much time, labour and financial investment, they have passed the point of no return and the solution does not fit in with the problem,” Bell told us. 

With the CivTech process encouraging start-ups to explore the extent of the problems that the government faces through collaborative sessions and extended exploration, Bell highlighted that this approach enabled his team to work directly with the people at the frontline of the problem, a connection between service and service user that was bridged through the constant deployment of prototypes, aided by HelpFirst’s machine learning engineer, who is also a former caseworker. 

Praising CivTech, the Scottish GovTech accelerator, Bell explained that by publicising the problem for startups like HelpFirst to see and providing time and resources for experimentation with affordable pilot schemes, they avoid forcing an expensive and potentially inadequate solution.

“It is hard to understand the full requirements for a new system in the abstract. The first articulation of a solution is most likely wrong. The beauty of the CivTech process is it allows us to explore multiple solutions. We build rapid prototypes and get them into the hands of the users. By seeing how the different prototype succeed - and where they fail - we build our intuition about the problem space, and navigate to a better solution. That exploratory approach isn’t possible with standard IT procurement,’ said Bell.

Compassion in user-centric design

Bradbury, who left the consulting juggernaut McKinsey for the Tech4Good startup world, joined Bell later in this journey, centralising the values of compassion, integrity and craftsmanship at the core of the HelpFirst business model. 

Whilst the GenAI hype is undeniable, the implications of it as a technology are still not fully understood. Cognisant of this, Bradbury’s role as COO is rooted in nurturing HelpFirst’s business values. Compassion is embedded into the design process, through understanding how the technology serves or inhibits the stakeholders involved in the technology.

Inheriting the craftsmanship approach from CivTech, HelpFirst ensures that the unique needs of the communities being served are represented in the design and final use of the product. 

In terms of integrity, Bradbury elaborated that this means doing the right thing even when no one is looking at the business and holding it accountable for its mistakes.

“Sharing what goes wrong and what is going well in the design process is a key part of how we practise integrity.” The fast development of prototypes, allowing failure to be seen, communicated and ameliorated is part of instituting integrity within the design process, and mitigates the “sledgehammer” propensity of GenAI. 

Ultimately though, the exposure of public services to the Covid Pandemic and the rise in the cost of living has been a wake up call to the need for radical transformation, and as Bell was a first hand witness of, the status quo sometimes comes with its own awful level of risk.

HelpFirst has shown that with the proper interaction, inclusion and iteration, GenAI can support meaningful and impactful solutions for both service providers and the vulnerable citizens they serve.