By Oliver Large, Policy Advisor, Tony Blair Institute for Global Change
Two very different job search experiences play out in Britain today. One job seeker refreshes her Universal Credit portal at 2am, hoping for a ten-minute slot with an overworked work-coach.
Down the street, a LinkedIn chatbot redrafts her neighbour’s CV, navigates him to his desired job that he “described” in a sentence, identifies the compatibility with his skills profile, and lands him an interview in under twenty seconds. While the labour market is racing towards the future, the state is stuck in the 20th century.
Britain’s employability architecture is drifting behind the labour market it services. At the centre of this is the Job Centre Plus (JCP). Pockets of excellence exist, but the data tells a stark story: of 2.5 million people unemployed or low earners, 60 per cent have been in receipt of UC for more than a year and 250,000 people for five years or more1. Fewer than 1 per cent of those on the higher rate of incapacity benefit move off the benefit each month2.
TODAY’S JCP EMPLOYABILITY SERVICES CAN’T SCALE BECAUSE IT IS BUREAUCRATIC, LABOUR INTENSIVE AND UNRESPONSIVE TO CLAIMANTS
This isn’t the fault of work coaches. It’s the result of a system that is bureaucratic, labour intensive and unresponsive to claimant need or preference. Good quality services are difficult to scale, even with significant investment.
Consider the pressure on staff: in 2024, JCPs employed 2,100 fewer work coaches than required, and 57 per cent of work coaches reduced the support they provided because caseloads were too high3. Much of what work coaches do – worth around £350 million per year in labour time – is spent on monitoring compliance for Universal Credit rules, time that could be used to help people find and progress in work. And while there is a heavy reliance on face-to-face meetings, job centres are unevenly spread across the country, which leaves many areas underserved. Faced with time constraints, coaches are more likely to default to the fastest route – usually, “any job at any cost.”
As a result of this operating model, the JCP has limited capacity to provide support to anyone or for anything else. Millions are excluded from quality support altogether, often labelled “too difficult to help,” despite research showing 20 per cent of incapacity benefit claimants want employment4. This is less a feature of the system and more a symptom of an operating model that cannot personalise support at scale.
Most support also stops on entry into the workplace. Disadvantaged groups, including disabled people, have a higher chance of falling out of work without adequate support5. In addition, the churn in the labour market will increase as AI automates and augments roles. The Tony Blair Institute for Global Change’s (TBI) analysis shows between 60,000 and 275,000 jobs per year could be removed, roughly one job every six minutes6. More jobs will be added, but this means workers need consistent access to upskilling, retraining, advice and networks throughout their careers – yet disadvantaged groups have less access to these opportunities.
Finally, employers are routinely not engaged. Only 14 per cent of employers engage with the Job Centre Plus7 – and just eight per cent of those that engage trust it to find them the right candidate. Improving these relationships will take time and money, with no guarantee of success.
In short, the current JCP model of employability services overburdens work coaches, limits access and quality for claimants, and fails to adequately engage employers. Resolving all these in the current operating model means spending billions more – an option that is neither politically nor financially possible.
AI MAKES POSSIBLE A NEW APPROACH TO EMPLOYABILITY SERVICES
AI presents a different path. It can both boost efficiency and unlock a new vision of service delivery.
First, AI can create efficiencies within existing workflows. According to analysis by TBI, work coaches alone could free up 45.9 per cent of time using AI tools, equivalent to a productivity gain of close to £295 million a year8.
The DWP has started to experiment with its potential: it has launched a £10 million initiative, Nexus AI, to build tools to improve call triaging, support work coaches to provide advice, and identify vulnerable claimants from their correspondence. In addition, the Department for Science and Technology are scanning the potential use of AI Agents in careers guidance9.
Second, these efficiencies also provide the capacity to reimagine what the service – and its staff – do. If AI tools handle routine admin (or other tasks like monitoring), Work Coaches can focus on complex, face-to-face support. For job seekers, this means resolving the obstacles to employment quicker and better.
Crucially, the productivity gains from AI adoption does not have to mean job losses. The DWP controls how to use these time savings – and should do so to build capacity for more human centred, relationship-based support. But first, it needs a radical vision of employability services in 2030, underpinned by technology and AI, to address the size and urgency of the problem.
AN EMPLOYMENT COMPANION FOR EVERY JOB SEEKER BY 2030
Today’s employability services limit support to a small group of claimants, under-prepares job-seekers and under-serves employers in the labour market. But AI opens up the possibility of a universal service – low barriers to entry, high quality, personalised throughout a person’s working life.
TBI has proposed giving every citizen access to a “Digital Employability Assistant” – outlined in Governing in the Age of AI: Reimaging the UK Department for Work and Pensions10. This Assistant would do several things.
One profile – based on an app or phone, the assistant would store a canonical version of the claimant’s CV, link to a lifelong learner ID (which holds their data about their education credentials and skills they have acquired through life).
Smart triage – based on the information held on their profile, the assistant would perform an initial virtual assessment to create an adaptive claimant commitment.
Employment springboard – AI models would analyse each person’s skills, preferences and local labour-market data to suggest suitable jobs, training and offer practical job-search help (such as tailored CVs, mock interviews and activity tracking).
These are not abstract ideas. France’s La Bonne Boite scrapes recruitment data from millions of companies and matches candidates to future vacancies. The job platform has four million active users (compared to 1.7 million users of job centres in the UK), a 70-80 per cent accuracy rating in predicting vacancies, and an 18 per cent increase in the effectiveness of getting people into jobs compared to those who do not use the platform11.
With the functions described above, the digital employment assistant can become an employment companion – supporting decisions about personal development, progression, and moving on throughout the claimants’ career. Citizens find jobs they can stay and grow in – not just any job that closes a claim.
The real breakthrough is access. AI never sleeps. Job seekers can engage with support whenever it suits them – critical for those juggling care or shift work. AI has already been used in commercial customer services settings to reduce call deflections by 80 per cent and contact resolution time by 50 per cent12 13.
In addition, AI’s near-zero marginal costs allows support to be extended to everyone – those within and beyond the benefit system, and those currently deemed too difficult to support. Some of the claimants who would traditionally be expected to go to the JCP could drop routine face-to-face appointments, freeing coaches for complex cases and those who prefer to talk to a human, who would still get longer in-person sessions backed by AI-enabled guidance.
Additional support could come from setting up virtual “career clinics” staffed remotely. In addition, where data shows rising need, temporary pop-up hubs in shared community venues could be created, allowing most permanent job-centre buildings to be phased out while widening access to employability support.
For employers, the platform reverses the information flow: it could push curated talent lists, matching employers to people with specific skills or encouraging engagement with bespoke programmes or pre-employment opportunities like traineeships. This would help to de-risk the application process for employers. And, freed from cold-calling firms to locate vacancies, the JCP can shift time and energy towards supporting candidates to maximise their chances.
Of course, this vision comes with risks. If poorly designed, AI tools could embed bias or reduce personal agency. For instance, if data is incomplete or has poor labelling, individuals could be penalised or miss out on relevant job opportunities. That’s why the Digital Employment Assistant (DEA) must be designed transparently, fairly, and with human oversight. This includes adopting an earned autonomy approach to implementing AI that TBI have outlined in previous reports14, and embedding the ability for citizens to speak to a human into all user journeys.
Human advisers, counsellors, and mentors will remain crucial in providing certain kinds of support, including emotional, ethical, and contextual support necessary for long-term success. The current system offers little room for this kind of support. Access to services should also remain in-person for a minority who cannot access digital tools. Our position at TBI is that using AI in the way described above is the only way to build the capacity to bring humanity back to the job search – away from tick boxing and monitoring to a focus on human flourishing.
The government faces a choice: keep managing a limited service that fails claimants, work coaches, and employers. Or with the right vision, and underpinned by AI capabilities, it could give every citizen an employment companion that turns opportunity into lasting work.
REFERENCES
- TBI analysis of data from DWP Stat-Xplore
- https://institute.global/insights/politics-and-governance/reimagining-uk-department-for-work-and-pensions – TBI analysis
ABOUT THE AUTHOR
OLIVER LARGE | Policy Advisor | Tony Blair Institute for Global Change
Oliver Large is a Policy Advisor at the Tony Blair Institute for Global Change, specialising in the intersection of technology and public policy.
His work focuses on the transformative impact of artificial intelligence on public services and the economy, with research spanning the NHS, welfare systems, local government, and the future of work. Oliver has written policy papers on reimagining the UK’s Department for Work and Pensions, improving service navigation in the age of AI, and how governments should respond to the impact of AI on the labour market.
He has advised political leaders in Europe and beyond on how to navigate both the opportunities and risks of AI adoption.