Public services need clear thinking – not hype. But if we follow the media it can be difficult to distinguish one from the other. AI moved fast in 2026. Some of it went well, some of it less so:
- Healthcare – We have seen the world’s first vaccine developed by AI.
- Criminal Justice – The UK Police have launched PoliceAI to speed up investigations and assist officers as they fight crime.
- Security – An AI agent breaking out of its controlled environment and hacking into a competitor’s servers.
Underneath these headlines, there are three trends which matter most for how we’re working in the public sector:
1. AI Agents
There has been a gradual shift from chat-bots answering questions to agents which can perform actions. This means that AI is starting to cross the line from supporting knowledge work to participating in workflows.
The efficiency gains are real, but agents are not the right fit for every problem. Deterministic workflows which have clearly defined inputs and outputs (such as processing planning applications) may be better served by looking at more traditional automation; and some domains are too critical to sub-contract out to any system, however advanced.
2. Multi-Modal AI is the New Normal
Gone are the early days of ChatGPT when we were simply putting in text prompts into a chat bot and expecting a text answer. Since 2023, AI models have become increasingly sophisticated in interpreting and manipulating other forms of media, including: documents, images, audio, code and video. This has accelerated the integration of AI with other tools and systems, allowing it to interact with a far wider variety of information sources.
Most workplace information is unstructured: being contained in dashboards, PDFs, maps, meeting recordings and hand-written notes. A lot of knowledge work up to this point has involved translating materials from one format into another. As multi-modal AI begins to accurately complete this work, the value of the worker shifts. Where some jobs may disappear, others will be transformed or created as workers transition from being content creators to content curators.
3. Model Development is Outpacing Human Skills
Anyone working with technology can feel like they’re behind the curve at times, but with the pace of AI development being so rapid, it’s often those who are working with it most who feel they can’t keep up. This applies as much to those building neural networks as to admin apprentices just starting their careers. In the public sector, we are not the ones developing this technology so there are always going to be ways in which we are behind.
But being behind on the technology is not the same as being behind on the application. These tools can perform a wide variety of tasks, but often in a shallow way: They can produce outputs on whatever you ask, but often these outputs cannot be fully trusted (however convincing they may sound). We can still be innovators by applying AI tools to our own domains. What does good AI in HR look like, or good AI in public health? It is only those working in these fields that can answer these questions and develop these solutions.
What this means for SODA
The Suffolk Office of Data & Analytics (SODA) has been working for the past few years to develop and upskill the public sector in Suffolk with AI. This has included:
- Delivering local and national training
- Sharing knowledge and learning through the Artificial Intelligence Teams channel
- Organising, hosting and running Suffolk’s first AI in Data & Analytics conference
- Developing agents in Copilot
More recently, we have been exploring how AI tools might transform the analytical process.
As AI becomes increasingly capable of doing analytical tasks, being skilled in AI will be as important as other technical skills we use every day. This blog is intended to be the first in a line of articles, commenting on how AI applies to data and analytics. It’s not about chasing all the latest trends and tools, but more about helping us all to have confidence to ask where AI can create real value and identifying what safeguards are needed to use it responsibly.
AI is already moving into the tools, systems and workflows that public-sector organisations use every day. The challenge now is not simply to keep up with the technology, but to develop the critical thinking, governance and evidence needed to use it well – knowing when an output is wrong, when we should avoid automation and when a confident answer deserves a second look. That is where SODA InnovAItion will focus: not on the hype, but on what AI means in practice for public services.
Steve Parsons
Advanced Analyst & Researcher
Suffolk Office of Data & Analytics