Releasing capacity across government: Where AI can deliver the greatest impact

The government faces increasing demand for public services, budget pressures and workforce constraints. Whilst much of the debate on AI has focused on automation, our analysis of >1,200 roles across the Civil Service¹ uncovered three key findings:

1. The AI opportunity extends across the entire workforce

AI is often associated with administrative and back-office work. However, our analysis found an average of 20 AI opportunities per role across all Government departments and agencies. And crucially the potential impact does not decrease with seniority.

This is not because AI can replace leadership judgement. Rather, senior roles (including at Grade 6 and SCS levels) spend more time creating, synthesising and reviewing information, producing briefings, developing policy and supporting decisions - activities where AI can provide significant assistance.  Individual roles in Research, Science & Analysis, Policy & Strategy and Digital, Data & Technology job families also show large potential impacts.

The implication is clear: AI should be viewed as an organisation-wide strategic capability, not a niche technology initiative focussed on operations.

2. The greatest AI opportunity is augmentation, not automation

The debate around AI often centres on workforce replacement. Our analysis suggests the reality is very different.

For most roles, there was a predominance of augmentation rather than automation. The average share of augmentation use cases ranges from 58% for the Prison and Probation service (e.g. 42% automation) to 62% for the Department of Education (38% of automation). (N.B. The chart below shows the percentage of augmentation minus the percentage of automation).

The greatest benefits come from helping staff analyse information, develop first drafts, produce briefings, conduct research and accelerate routine activities.

True end-to-end automation opportunities do exist, particularly in high-volume transactional functions (e.g. contact centre and back-office processing roles), but these represent a relatively small proportion of the use cases identified.

For most organisations, the challenge is therefore not reducing headcount but enabling employees to spend more time on decision-making, improving delivery and providing frontline services to the public.

3. Most departments and agencies already have access to the technology for AI

A common assumption is that successful use of AI requires more technology and bespoke AI solutions. Our findings suggest otherwise.

87% per cent of identified opportunities could be addressed using general-purpose AI tools (e.g. Copilot, Claude, Gemini) rather than specialist platforms.  

Beyond areas like cyber-security and intelligence where custom-built models rightfully earn their place, Science and Digital are the main job families where specific AI might be needed. Even in those areas more than half the opportunities can use Generic AI.  The highest share of coverage for Generic AI are within Corporate support, Communications, and Operational Delivery (some of the largest areas of the public sector workforce).

Most government departments and agencies have access to a general-purpose tool (often Microsoft Copilot). Technology availability is therefore becoming less of a constraint.  The real challenge is therefore how to encourage adjustments to ways of working and support greater adoption.

The leadership challenge

Many departments and agencies remain focused on pilots and productivity improvements for individuals. Having provided access to an AI platform, they are hoping individuals will innovate and change how they work independently.  Whilst a tool like Microsoft Copilot can save 30 minutes per day, individual productivity is just the tip of the iceberg, a fraction of the potential opportunity.

The larger prize comes from rethinking how teams operate and daily processes run end-to-end. This could include streamlining policy development and briefing processes or accelerating analysis and knowledge-sharing. It targets a reduction of administrative effort or duplications and redesign of operational processes.  This requires leadership at the top and from managers.

AI is not primarily a technology transformation. It is a workforce transformation enabled by technology.