From pilot to productionised AI at scale: Tips for public sector leaders

Ali Bovis at Version 1 discusses why artificial intelligence presents a significant economic opportunity for the UK and offers valuable advice for leaders in the public sector

Artificial intelligence (AI) is a significant economic opportunity for the UK, with Treasury figures indicating it could add £400 billion to the economy by 2030. With the government setting ambitious targets of 2% annual productivity improvement and 5% efficiency savings across departments, AI implementation has never been more urgent.

Despite direction from the AI Action Plan and AI Playbook, only about 20% of government AI initiatives move beyond pilots. Recent government publications highlight persistent challenges: out-of-date technology, data quality, skills shortages, and inadequate governance as key contributing factors.

The £45 billion opportunity

Government analysis suggests AI and digital transformation could deliver £45 billion or more in annual productivity and efficiency savings across the public sector – £36 billion from automating delivery, £4 billion from online services, and £6 billion from reducing fraud and error. This aligns with the government’s view that “No person’s substantive time should be spent on a task where digital or AI can do it better, quicker and to the same high quality and standard.”

In our experience, effective AI implementations, and those that deliver the multi-billion pound benefits, are not revolutionary headline-grabbing moonshots. They often come from augmenting the capabilities of frontline public sector workers with AI – nurses, social workers, teachers, and many others – empowering them to prioritise delivering care and having meaningful citizen interactions. The solutions Version 1 has already deployed demonstrate that even single-use cases can return multi-million-pound efficiency benefits.

The challenge is how to identify, build and deliver material use cases in the most effective way and into daily production use.

Bridging the implementation gap

Drawing on experience implementing AI across dozens of public sector organisations, we have developed a proven approach to co-creating business value and moving beyond experimentation. We use four foundational pillars to address the gaps:

• Approval:
  • Define specific use cases with clear business outcomes, conduct thorough legal and privacy assessments, and secure stakeholder buy-in. Our experience shows that initiatives with senior sponsorship and strategic alignment are three times more likely to reach production.
• Governance and accountability:
  • To satisfy Cabinet Office controls, establish a departmental AI Assurance Board with cross-functional representation and implement a tiered governance approach matching oversight to risk levels. We have found that with suitable risk-based controls, effective governance accelerates, not impedes adoption.
• Upskilling for success:
  • Start with basic AI training for all staff to create a common language and reduce resistance. Follow with role-based training pathways: e.g. technical teams need hands-on AI coding experience (in our experience, saving 30-60 minutes per developer daily). Finally, establish AI champions to accelerate peer learning, which our data shows doubles adoption rates compared to centralised training alone.
• Measurement and improvement:
  • Set metrics for success that link directly to organisational objectives and continuously evaluate AI implementations. This is essential for demonstrating progress toward the government’s efficiency targets.

Real-world impact

Tackling fraud and error:

Our work with a key government department demonstrates AI’s impact on public finances. Working with partners, a proactive fraud prevention strategy has been implemented using AI to analyse claims in real time. This solution has already contributed to savings of over £1.3 billion, with significant additional contributions projected by 2027/28. This directly supports the government’s fraud reduction targets.

Improving experiences:

Our collaboration with the Children and Family Court Advisory Service (Cafcass) demonstrates how AI delivers tangible benefits to vulnerable citizens. Serving 140,000 children annually, Cafcass advisers struggled to effectively personalise their 80,000 monthly communications. Our Scribe solution now uses AI to tailor letters to each child’s reading age, generate audio alternatives, and provide translations – significantly improving engagement for children navigating complex court proceedings. This implementation enhances the citizen experience and returns valuable time to frontline staff previously spent on administrative tasks.

Reducing time and duplication:

Some environmental and agriculture departments are already innovating with multi-modal AI models. For example, using audio and visual AI to assist automated form input promises to reduce livestock inspection times significantly.

We have also supported using AI- assisted compliance checks to increase the number of vehicle inspections when crossing borders, saving over 1,000 hours per month. Similar solutions are being piloted with medical practitioners, where ambient voice-based clinical notes can be generated during patient consultations, saving valuable clinician time in trials in several hospitals.

Reaching the production line

The Committee of Public Accounts highlighted that “realising the benefits of AI across the public sector will require strong leadership”. We employ a challenge-first methodology that begins with the problem statement and expected value. Our approach is people and not technology-centric, and we assist in driving meaningful adoption of AI solutions using the ADKAR (Awareness, Desire, Knowledge, Ability, Reinforcement) framework-based change management approach.

Importantly, given the potential human impact, we believe in AI being a collaborative partner with people, rather than a replacement. The “human in the loop” model with retained roles as trainers, validators, orchestrators, collaborators, and final decision-makers – builds confidence and reduces resistance.

Five actionable steps to production

1. Conduct an AI readiness assessment:
   • Start with targeted assessments at a specific process/use case, or project level rather than attempting department-wide change. Focus on specific capabilities – data quality in one area, skills in another – creating a mosaic of insights that informs your direction.
2. Start with high-value, low-complexity use cases:
   • Focus initial efforts on areas with clear business value, available data, and manageable complexity – create early successes to build momentum.
3. Establish a lightweight governance process:
   • Implement appropriate guardrails without stifling innovation; start with core principles that evolve as your AI maturity increases.
4. Develop appropriately skilled cross-functional teams:
   • Combine technical expertise with domain knowledge and change management skills – diverse perspectives lead to more robust implementations. Team construct is important and not dependent on technology. Providing an opportunity to build AI experience is about aptitude and attitude, not technical experience.
5. Create a measurement framework:
   • Define clear success metrics that link directly to business outcomes, allowing you to demonstrate value and continuously improve.

From experimentation to transformation

The public sector stands at a critical juncture in its AI journey. The potential is enormous – financially and in transforming services for citizens and staff alike.

While government guidance continues to evolve, organisations need not wait for step-by-step instructional direction. By building a solid foundation – looking at skills, appropriate governance, addressing cultural resistance, and focusing on outcomes – public sector leaders can bridge the gap and benefit from AI’s £45 billion potential.

Please Note: This is a Commercial Profile

This work is licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International.