Artificial Intelligence has reshaped public engagement and consultation

Artificial Intelligence (AI) is already inside your consultation process, whether anyone signed off on it or not. The organisations that stay in control will be the ones who understood it before they leaned on it.

Getting started with AI is the easy part. Almost anyone can open a chatbot and start typing. The harder part, and the part most organisations are behind with, is staying in control of it once it’s in your practice.

Picture someone paddleboarding for the first time. Getting on the board takes no skill at all. But without a paddle, you’re not steering, you’re drifting. The wind decides where you go, not you. You’re still on the water, technically participating, but you’ve handed control to something else without quite meaning to.

That’s where a lot of engagement and consultation work currently sits with AI. Staff and respondents are already using it, often without anyone deciding that they should, and the gap between using AI and understanding it isn’t just a skills problem. It’s a governance and risk issue, and it needs treating as one.

AI is already mainstream, whether you planned for it or not

ChatGPT alone reaches over 16 million people a month in the UK, according to IAB.UK – the UKs industry body for digital advertising. Last week NHS England announced it is rolling Copilot out to over 500,000 clinicians and support staff. AI isn’t a novelty anymore. It’s mainstream, it’s in your building, and your respondents are already using it too.

That changes engagement and consultation work on both sides of the table.

Your staff are already using AI

Engagement and consultation professionals are quietly using AI to draft, analyse, test scenarios and check accessibility. Some of it is deliberate but a lot of it isn’t. People are experimenting, finding a use for AI that saves them an hour, without thinking too hard about how the tool behaves or what the consequences might be.

This is the part that concerns us. In our training sessions, we keep finding that people don’t know about the sycophancy problem. AI will agree with you even when you’re wrong. It will also fill in gaps when it doesn’t know something, unless you explicitly tell it not to. In public-facing work, that’s how misinformation could quietly end up in a briefing document or a consultation summary.

Over-confidence is a risk too. If staff can’t analyse data manually, how can they be expected to use AI to do it well? AI accelerates whatever’s already happening, whether that’s good practice or bad practice.

Consultation work depends on defensible methodology, human judgement and a clear evidence trail. Recommendations and decisions cannot be shaped by AI.

If you lead an engagement and/or consultation function, there are several questions you should be asking today:

• Where is AI already in our workflows (officially or not)?
• Do staff understand where these tools fail, not just where they help?
• Do we have evidence logs showing how AI was used?
• Are we confident every recommendation is still fully human?

A team that can’t answer these doesn’t have a paddle yet. They’re on the board, but the wind is in charge.

Your respondents are using AI too

While organisations work out their internal rules, the people you’re consulting are changing how they engage, often without telling you.

They’re drafting responses with it. Some people use AI to express themselves more clearly. Others use it to write the response for them. Either way, you’ll need a view on what to do with a wave of replies that read suspiciously similar. Is that really so different from a coordinated campaign handing out a response template?

They’re reading your documents through it too. Long or technical materials are increasingly summarised by AI before anyone reads the original. That might help accessibility, but it also means people may be reacting to someone else’s summary of your policy rather than the policy itself, which raises the stakes on making your original material genuinely accessible from the start.

And it’s sitting in on your meetings. Scribe-bots are already turning up to online engagement sessions. Worth deciding now what your rules are, and whether you allow it at all.

What this means for the law you already follow

The Gunning Principles haven’t changed but AI has changed some of the challenges. This is where understanding AI stops being optional and starts being your legal safeguard.

If respondents are engaging with an AI-generated summary rather than your actual document, are they really getting sufficient information? And if AI has had any hand in shaping how your team analysed responses or arrived at a recommendation, can you still show that conscientious consideration sat with a human, not AI? These aren’t new legal tests, but AI has made them harder to evidence. An organisation that hasn’t taken the time to understand how its tools behave is not well placed to answer either question with confidence.

Where AI actually earns its place: analysis

Analysing large volumes of response data is expensive, slow, and a genuine pain point for most engagement and consultation teams, which is exactly why people are reaching for AI here.

There is certainly a role for AI. The UK Government’s Incubator for AI built Consult specifically for this: in 2025 it analysed more than fifty thousand responses to the Independent Water Commission review in two hours, with experts verifying the results.

In terms of data, we’ve had success using our own tenant-protected Copilot to support data analysis, with our robust oversight. We’ve used products with integrated AI analysis tools, and we’ve had demos of other AI analysis products in development: this is a genuinely exciting area to keep an eye on.

One principle will hold regardless of how sophisticated the AI tools get. Whilst AI can speed up the bit before a decision is made, it cannot make the decision. Back to our paddleboard metaphor, AI can help you paddle faster. It still can’t tell you where you should be going. That’s the part that has to stay human, and it’s also the part that gets harder to defend if you don’t fully understand what the tool did to get you there.

Help is already on the way

Open Voice Lab has built foundational and advanced training to step engagement and consultation teams through exactly these challenges, and we’re excited that it’s now part of the Consultation Institute’s training programme. Beyond training, we can also support with strategy and the practical work of embracing AI responsibly across your engagement function.

The organisations that will be fine

Research from the Alan Turing Institute found that 72% of people say regulation makes them trust AI more. People aren’t asking for AI to vanish from public life. They’re asking for proof someone’s keeping an eye on it.

The organisations that come through this well will be the ones who understood AI before they deployed it, not after, with clear governance. Ideally that plan gets built with the communities and stakeholders affected by it, not just handed down to them.

AI isn’t replacing engagement and consultation work, but it is changing the room it happens in. The skills that result in good, defensible, engagement and consultation work haven’t gone anywhere. They matter now more than ever.

The question isn’t whether AI belongs in your practice. It already does. The question is whether you understand it well enough to be the one steering. If you’re working out where to start, that’s exactly what our foundational and advanced AI training is built for, now part of the Consultation Institute’s training programme.

---

How tCI Can Help

Advice and Guidance
A tCI faculty member will work alongside you to support the development of your decisions and engagement approach. We provide independent, constructive advice at critical stages, helping you strengthen stakeholder mapping, test communication strategies, and plan robust post-decision engagement. Our role is to act as a critical friend, offering practical recommendations grounded in consultation law and good practice that build confidence in your process.

Risk Assessment
Early identification of legal, political or reputational risks in your engagement approach. Using tCI’s five-risk methodology, we spot gaps before challenge arises, helping you strengthen stakeholder communication and demonstrate procedural fairness from the outset.

Executive Briefings
Concise updates for senior leaders on consultation law, engagement duties and post-decision risks. Helps boards and leadership teams make confident, defensible decisions when under pressure, with clear guidance on what good engagement looks like after difficult choices are made.

Whether you’re preparing for a high stakes service change or building defensible evidence for complex decisions, we can help.

Contact tCI: hello@consultationinstitute.org