3 phases of public sector AI adoption

In the 12 months since the Government Digital Service launched its Artificial Intelligence Playbook for the UK Government, the conversation within the UK public sector has shifted. The initial excitement about the potential of AI to transform public services and improve operational efficiencies has been replaced with the realities of real-world implementation.

At Equal Experts, partnering alongside major government departments like Defra and another large government department has shown us that successful AI adoption is more than a single leap. Instead, it’s an evolutionary journey impacting individuals, delivery teams and eventually the entire department.

In this new blog series, we’ll explore each stage in the journey in depth, including what it typically looks like, some of the successes we’ve seen and how public sector organisations can consistently move forward with AI.

The three phases of AI adoption

In our work supporting government and public sector organisations to introduce AI into their core delivery capabilities, we’ve seen three distinct phases emerge:

1. Innovation phase: Through small experiments, proofs of concepts (POCs), and simply having a play with new tools and approaches, organisations start to understand more about AI and its potential.
2. Industrialisation phase: Organisations begin to move from isolated POCs into scaled, production-ready services, with an increasing focus on centralised governance, cost management and strategic prioritisation.
3. Organisational coordination phase: AI becomes a cross-departmental capability with consistent standards, shared platforms and dedicated leadership roles such as Chief AI Officers.

Defining the innovation phase

As a new frontier of technology, early AI adoption is characterised by curiosity, with high-energy, low-friction activities. Organisations, especially those in the highly-regulated and citizen-focused public sector, often start out with a few ideas and smaller teams to test and learn while limiting the blast radius for any early errors.

Typical focus areas may include:

• Rapid prototyping: Testing AI’s ability to augment the software delivery lifecycle (SDLC), from discovery and design to testing and maintenance.
• Tool exploration: Experimenting with AI assistants, AI agents and AI capabilities within existing tools, such as Microsoft Copilot and Github Copilot
• Small experiments: Small but enthusiastic groups working in “sandbox” environments to learn safely.

Different approaches shaped by the organisation

Just as there is no one single AI experiment that every organisation starts with, there is no single blueprint to innovation within the public sector. Each organisation is unique, with different infrastructures, leaderships, data sensitivities, risk attitudes and security postures all playing a role in shaping AI decisions.

Delivering value in the Innovation phase

It’s easy for organisations to fall into the mindset that only significant, large-scale AI programmes can make a difference. But we’ve seen public sector organisations reap significant rewards in this early phase by turning early demos into tangible delivery capabilities with real value. Many organisations see opportunities to create mini-products that previously would not have been cost effective to build.

Defra: AI to enhance software delivery

Equal Experts partnered with the Department for Environment, Food and Rural Affairs (Defra) to explore how GenAI could responsibly enhance software delivery. From the outset, Defra chose a hands-on approach to AI adoption, writing production grade code with AI, and designing a “best-fit” team structure around AI capabilities. It resulted in the development of Defra’s AI SDLC Playbook, a guide to integrating AI into software engineering workflows responsibly, without compromising quality. Read more about our early AI innovation work with Defra.

Removing bottlenecks with an AI-enabled domain API approach

A large government department’s API estate is complex with an estimated 2,000 APIs in use, many of which have overlapping or duplicated business concepts, leading to high change costs and slow delivery. Equal Experts worked with the department to test, through real delivery, whether AI could materially reduce time, cost and risk in a domain API approach.

During the proof-of-concept, we discovered that AI can rapidly analyse API specifications, documentation, and code to identify domain entities, propose consolidated domain APIs, and assess migration impact across services. We also successfully deployed AI-generated APIs and services within the department’s existing environments using established AI governance and security guardrails. A AI assisted Domain API Playbook was also created and can be used across the organisation.

Results include:

• End-to-end service delivery reduced to 2-3 weeks with AI, compared to 6-8 weeks traditionally
• API and service scaffolding reduced 1–2 hours with AI, versus days to weeks manually
• Domain discovery across APIs was reduced with tens of API specifications analysed and grouped in days with AI, compared to weeks to months using manual review
• Multiple repositories and services analysed in hours to days with AI, compared to weeks of manual investigation
• If adopted at scale the cost of developing an API would reduce from hundreds of thousands to tens of thousands.

This approach removed the primary cost and time bottleneck of early analysis and design, while operating fully within existing security and governance frameworks. It enables faster delivery with smaller teams and no increase in headcount, reduces duplication and long-term run and change costs, accelerates policy and service change through reusable domain APIs, and transforms central integration functions from delivery bottlenecks into platform enablers. While human validation remains essential, with all outputs validated by departmental architects, the effort required is significantly reduced, ensuring speed without loss of control.

The innovation trap and how to escape it

The greatest risk in early AI adoption is getting stuck in the innovation phase. The excitement of creating cool demos and trying new tools can quickly lead to an endless loop of experiments that never reach production and therefore never add value.

Pilots often fail to deliver because they are:

• Disconnected: Ideas that don’t align with the department’s wider strategy.
• Unmeasured: There is no framework to track the actual impact or return on investment.
• Isolated: There is no repeatable delivery process others can follow, with success depending on one or two AI champions.

Breaking out of this “pilot purgatory” requires the excitement of experimentation to be met with engineering and organisational discipline. To successfully transition out of the innovation phase and into industrialisation, organisations need to focus on:

1. Strong sponsorship: Find a leader with the authority to push beyond the sandbox into production.
2. Standardised governance: Rethinking existing business models and processes, focusing on how outcomes can be achieved differently using AI.
3. Real-world application: Moving AI onto actual work stories and maintenance tasks rather than just side projects.

Looking ahead

This is the first in our blog series focused on AI adoption in the public sector. In the next blog in this series, we’ll look in more detail at the industrialisation phase, including the scaling patterns, metrics, and anti-patterns to avoid when rolling AI out across multiple teams.

If your organisation is stuck in the first phase of AI adoption, we can help. Contact Equal Experts to find out how we can support your transition out of the innovation phase and into industrialisation.

About the author

Kev Gray is a Client Principal at Equal Experts UK, bringing over 25 years of experience in service and digital transformation for the public sector. He drives change and supports UK government teams to modernise legacy systems, embed data-driven decision-making, and deliver improved service outcomes.

Before joining Equal Experts, Kev held senior roles driving service delivery and stakeholder engagement across both public and private sectors. He is committed to building sustainable digital capability within government teams and fostering long-term, value-led collaboration. Connect with Kev Gray on LinkedIn.