Former chief executive of the Local Council Roads Innovation Group (LCRIG), Paula Claytonsmith, has announced that she has returned to work in highways and is already leading a key project with the highly influential and ADEPT-affiliated Future Highways Research Group (FHRG).
After stepping down from her role at LCRIG, Ms Claytonsmith took a short break last year to 'focus on her health'.
Announcing her return to work on social media, Ms Claytonsmith thanked the sector for its 'sheer generosity and kindness' and the many messages of support she had received. She added that it was 'brilliant to now be back' but stressed that she would be pacing herself.
On her return, Ms Claytonsmith, one of the most influential and well-connected figures in the UK roads sector, has already been named a 'research theme leader' by the FHRG.
She will be looking into 'assignable and agentic AI', including looking into costs and system-level risks for the sector.
Ms Claytonsmith revealed that she would be working with a select number of authorities while researching the vital issue. A progress report is expected in September this year, with next step recommendations.
She added that there was 'more to come' on the future of the research programme, with updates expected shortly including news about Tier One involvement.
Under the project, agentic and assignable AI is defined as AI systems that can be delegated defined tasks and objectives, even operating semi-autonomously within guardrails.
The AI can also interact with other systems, data, and humans and escalate decisions or exceptions when thresholds are breached.
Initial briefing notes on the research programme, seen by Highways, outline potential costs and risks of the agentic AI technology, including upfront investment, model bias and brittleness under abnormal conditions, the sector becoming over-reliant on automated recommendations as well as cybersecurity and data governance risks.
Potential system-level risks for local highway authorities identified from the technology also include 'optimisation that unintentionally degrades safety, loss of situational awareness through abstraction, fragmentation of accountability and misalignment with statutory and regulatory duties'.
