Christian Graham10 Mar 2026
When people talk about AI in the charity sector, the conversation usually lands on chatbots, content generation or data analysis. Useful stuff, but familiar. What gets far less attention is something more unsettling and potentially more powerful: simulation.
AI can now role-play. It can adopt personas, argue positions, anticipate objections and stress-test ideas in ways that feel genuinely useful and occasionally unnerving. At Friends of the Earth, I've been experimenting with this capability. The results have made me think differently about what's coming.
Virtual focus groups
One of my first experiments was simple: I asked an AI to simulate a focus group of different stakeholder types responding to an idea about installing domestic solar. A sceptical middle-aged businessman. An older nurse. A climate-anxious teenager. All living in east London.
The responses weren't perfect, but they were useful. They surfaced objections I hadn't considered and if you tweaked the personas, the responses changed too. For example, the nurse persona was initially keen on solar but became much more circumspect when it was revealed she was renting her home and didn’t think her landlord would be keen.
A valuable prototyping tool
This approach can help us rehearse difficult conversations before having them for real. They’ll let us iterate faster than waiting weeks for actual user research.
I’m not proposing this as a replacement for genuine engagement. Nothing substitutes for listening to real people. But as a rapid prototyping tool, it's valuable. You can test ten framings of a message in an afternoon, refine the strongest ones, then take those into real-world testing.
The darker side
Here's where it gets uncomfortable. If we can simulate stakeholder responses, so can everyone else.
Governments could use AI to simulate how NGOs will react to policy announcements and then design communications specifically to neutralise our objections before we've even made them. Corporations could rehearse crisis responses against AI-simulated activists, finding the precise language that deflects criticism without conceding anything.
This isn't speculation. Simulation and scenario-planning are already standard practice in political communications and corporate PR. AI just makes it faster, cheaper and more accessible. The asymmetry that civil society has often relied upon - that we're more nimble than lumbering institutions - may be eroding.
Staying ahead
For campaigning organisations, the strategic implications are significant. If our opponents can simulate us, we need to get better at being unpredictable or at least at understanding how we're being modelled. If AI can rehearse our arguments before we make them, we need to find new arguments, new framings, new coalitions that aren't in the training data.
And we need to experiment with these tools ourselves. Not to replace human creativity, but to augment it. To test ideas faster. To anticipate countermoves. To prepare for conversations that matter.
At Friends of the Earth, we're still early in this work. I've used AI personas to stress-test programme designs, to explore how different audiences might receive our messaging and to rehearse stakeholder conversations. The results have been mixed but promising. But useful enough to keep exploring.
What comes next
Simulation is one of AI's least discussed and most consequential capabilities. It doesn't generate content so much as generate possibility spaces, letting you explore how things might unfold before committing to a path.
The organisations that understand this early will have an advantage. The ones that don't may find themselves outmanoeuvred by opponents who've already rehearsed every move.
I'd rather be in the first group.