In 2025, if you work in public sector comms you’re likely to be worried about AI while also using it while not admitting you are.
You are also likely to be using AI tools secretly on your own devices.
But if you are using AI, you are generating ideas rather than generating audio, video or images.
These are the findings of a poll of more than communicators I conducted in early summer 2025.
It’s a set of figures that exposes a sector being asked to meet Sillicon Valley expectations with depleted analogue resources.
Prime Minister Keir Starmer has talked about AI being the challenge of the generation. UK Government have also talked about £45 billion savings being powered by AI despite Gartner still placing AI in the ‘peak of inflated expectations’ section of the hype cycle.
Right now, the challenge of the century is being put on hold with sector reorganisation often a far higher priority.
Let’s go through the numbers.
The sector is both worried and excited
Firstly, the public sector is torn between being worried and being excited. Almost three quarters told the survey that they felt both worried and excited by AI. This is understandable. New technology brings opportunities as well as downsides. With AI, we’ve not even begun to work through them.
Worryingly, much of the sector is operating without a policy
UK Government guidance in the AI playbook has set out 10 principles that can be used to use AI safely. Government Communications Service have also produced guidance around generative AI. If you are part of the UK Civil Service you are covered. However, while wise communicators can learn from these approaches there is no obligation for them to be enacted by arms length bodies, NHS, Police, Fire & Rescue or by devolved Governments.
Alarmingly, more than half of public sector people who took part in the poll said they had no corporate AI policy. Just 26 per cent said they had such a document to map out how AI will be used.
This represents a worrying gap in governance and leaves open the chance of some potentially catastrophic errors being made. An AI notetaker that is not secure being used to log a potentially sensitive meeting may be accessed by a bad actor. The serious case review being exposed to an unsafe notetaker has the power to be career ending as well as put people in harm’s way.
Much of the sector is not declaring how AI is being used
Transparency is a key word in many published AI policies in the public sector and also the third sector. Trust is built by being open, the GCS generative AI policy says. So does Friends of the Earth.
A lack of openness is storing up problems. Organisations have a chance to dictate when and where the story is first discussed. They have, as some would say, a chance to shape the narrative. That window is small and is gone for good when someone discovers that AI is being used and starts asking questions. Those questions are not unreasonable.
Just five per cent have a dedicated webpage disclosing what AI is used while less than one in ten disclose AI on every piece of content that’s produced with it. Just 11 per cent have had a conversation with stakeholders while 47 per cent say their organisation doesn’t admit they are using it.
‘We use AI ethically but we don’t declare it,’ caught my eye. One reading of this is that intentions are good. Another is that there is an unresolved struggle between what is the right thing to do.
Much of the sector are using AI on their own devices
The internet doesn’t care about a block, as the saying goes. It just moves around it like water moves around a boulder.
Anecdotally, people admit to using a tool like ChatGPT on their own devices because they are blocked from using them on work devices. This entirely mirrors early policies which blocked social media from public sector networks. Being charitable, perhaps bans are put in place with the best interests of the organisation at heart as an AI policy is missing.
Either way, people using AI tools at home are likely to be taking risks.
The additional comment ‘we don’t use AI… officially’ was enlightening in this section.
Most public sector comms are using AI to create ideas, not bad pictures
Interestingly, the top reason for using AI for public sector comms is to generate ideas with three quarters citing this as a purpose. Spell checking and grammar came in second at just less than half while a third generated press releases and just over two in 10 generate images.
So, it is the inspiration part of the process that AI is helping with as opposed to the hard content.
Within the additional comments, creating social media content appeared strongly. There were honourable mentions for using AI to summarise existing content as well as to make a start on creating text be that policy or comms plans. Transcription was also a frequent use.
The most popular tool is ChatGPT
Surprisingly, although Copilot is Microsoft’s AI tool for Office 365, it was not the most used tool in the public sector. This crown went to ChatGPT.
In the additional comments, there was slim pickings with occasional mentions of UK Civil Services’ in-house GCS Assist was mentioned in the comments but was not featured in the initial survey. I’d love to see the numbers for these in the coming months.
Elsewhere, there was not a lot of declared adventure with occasional references to DaVinci Resolve the video editing tool as well as other Adobe tools such as Photo Shop.
Conslusion
Walking to work this morning, I listened to a podcast where the presenter was interviewing an author who had wrote some poems. He told the author, a comedy writer and comedian, if he minded if he ran a poem through Google Notebook LM which then produced a podcast-style conversation to critique the work.
After the author’s indignance that this was possible the clip of the AI generated chat was played. There was much shock and surprise that this was possible.
I have a theory that these moments are getting faster and don’t last as long as they used to.
I’ve no doubt that I’ll look at this clip as an example of people in 2025 being amazed at something that will become something as straight forward as a web search.
It’s clear that in 2025, the public sector may have been told that AI is important but it is falling behind the rapid path it needs to take to do it well.
Methodology and prize winners
The survey was run in May 2025 and received 350 replies. A total of 291 were from the public sector which forms the basis of the results published.
As an incentive, a place on my ESSENTIAL AI FOR PUBLIC SECTOR COMMS and a box of Gower Cottage brownies was offered as part of prize draw to people who completed the survey. Everyone who completed the survey and indicated they wanted to take part were included in the prize draw. First prize of a training place and brownies was won in the random draw by Nick Moore of Leeds City Council. A box of brownies was also won by Kia Newland (Notts Hospice) and Anna Toone (Tamworth Council).
Thanks everyone who took part.
AI disclosure: I used Grammarly to check spelling and grammar in this post.
I deliver training to help you make sense of the changing landscape including ESSENTIAL AI FOR PUBLIC SECTOR COMMS.















