The shift from apps to agents marks a turning point for public sector technology, and success will depend on treating AI like a workforce, with defined roles, measurable outcomes and accountability built in from the start. Write Hitachi Solutions’ Jason Almeida and Jack Nutkins
Public sector organisations have spent years using the Microsoft Power Platform to modernise services. We are now entering a new phase. AI agents are active participants in workflows and not just passive tools. This shift requires a fundamental rethink of capability, accountability and workforce management.
The governance lessons learned from scaling low-code deployments apply directly to this new era.
From apps to agents
In the early days of low code, the focus was on building apps and automations quickly. Governance often followed later, sometimes much later. Many organisations found themselves with hundreds of flows and apps, unclear ownership, and growing operational risk.
Agents are different. Or at least, they force a different mindset.
Rather than thinking of agents as applications, it is more useful to think of them as digital workers. Unlike traditional software, agents are non-deterministic and act with a degree of autonomy, operating across systems and data. This means they behave more like humans than software, requiring an identity, a clear purpose, and defined boundaries.
Microsoft’s agent ecosystem increasingly reflects this shift, introducing concepts like agent identities, role-based access control and sponsorship to ensure every agent is accountable to a named individual.
This is a significant evolution. It addresses a key issue from the low-code era, where ownership was often ambiguous. In the agent era, ownership is designed in from the beginning.
Governance arrives earlier this time
One of the more encouraging shifts is how quickly governance is being considered and activated with tools like Agent 365. With Power Platform, governance frameworks often lag behind adoption. With agents, the gap is much smaller.
There is a growing recognition that governance is not a blocker to innovation, but an enabler. Particularly in government, where data sensitivity, sovereignty and accountability are critical, this matters. With the right processes and frameworks, governance can provide a safe on-ramp enabling makers while maintaining the necessary guardrails.
Agents also expose existing weaknesses. For years, organisations have relied on what might be called “security by obscurity”. Access was technically open, but practically hidden. Agents change that. In our experience, they surface information across datasets, sometimes in unexpected ways, making existing access issues visible almost immediately.
In that sense, agents are not creating entirely new risks. They are accelerating and amplifying ones that were already there.
The importance of value and intent
Another lesson carried forward from low code is the importance of clarity. In the Power Platform world, success often came from clearly defined use cases and measurable outcomes.
With agents, that discipline becomes even more important.
Because agents can do so much, there is a real risk of deploying them without a clear purpose. Unlike traditional automation, where outputs are predictable, agents can consume resources, interact with systems and influence decisions in more fluid ways.
You would not hire someone without a job description and KPIs to measure success. The same principle applies here. Defining what an agent is for, how success is measured and what “good” looks like is essential. Without that, organisations risk cost without value.
At the same time, the upside is significant. We are not talking about incremental efficiency gains. In our experience and based on Microsoft’s Work Trend Index, we are seeing step changes in productivity, far beyond what traditional automation delivered. That potential is exactly why governance needs to be designed in early, not bolted on later.
From shadow AI to governed enablement
If there is one piece of advice for public sector leaders, it is this. Assume AI is already in use across your organisation.
Just as low-code tools once appeared organically in departments, AI tools are now being adopted at pace. Whether it is Copilot, third-party tools or experimental platforms, usage is often happening before formal strategies are in place.
This creates a familiar but more complex challenge. Shadow IT becomes shadow AI.
The risks are not only technical but regulatory. Data sovereignty, in particular, remains a critical consideration for UK public sector organisations. Even within trusted ecosystems, there are nuances in how models operate and where data is processed.
The starting point, therefore, is visibility. Build an inventory of where AI is being used. Understand the scale and nature of that usage. From there, organisations can introduce guardrails that enable safe adoption rather than trying to shut it down entirely.
If the official route is too difficult, people will find another way.
The organisations that succeed will balance enablement and value realisation with concrete technical controls. Tools within Microsoft Purview including Data Classification, Data Loss Prevention, Security Posture Management ensure digital workers respect established security boundaries and prevent the accidental exposure of sovereign data.
Governance as talent management
In the agent era, governance becomes closer to talent management. Agents need roles, oversight and accountability. They need to be monitored, evaluated and, where necessary, constrained (or even fired!).
It is not about creating policy for its own sake. It is about building a system that can scale safely.
Moving forward
We will explore these themes in more detail through our upcoming PowerCAT events with Microsoft in London. Each session is designed to help public sector organisations move from experimentation to structured, scalable adoption. Follow the links below to register.
Jason Almeida is a solution architect with over 10 years of experience designing and implementing business applications that drive transformation through process optimisation, automation, and AI. With a passion for technology and how it can revolutionise the way people and organisations work, he partners closely with clients to understand their objectives and challenges, crafting solutions that deliver real, measurable outcomes.
Jack Nutkins, head of power platform at Hitachi Solutions, specialises in delivering large-scale enterprise solutions across both the public and private sectors. With a strong background in architecture, go-to-market strategy, and cross-practice leadership, he is reimagining how the Power Platform is viewed as a technology-enabler.
