Your strategy would benefit from a marketing director, a behavioral scientist and a policy analyst. You probably can’t afford any of them. And whatever you publish next — a website, a report, a press release — will be read by journalists, policymakers, farmers, council officers, funders and the public, with no easy way to ask them in advance how it lands.
AI can help with both. These models have been trained on millions of words written by those people, so you can use AI to simulate them and then talk to them. This guide walks through building one reusable persona and using it on a real piece of your work. It takes about twenty minutes.
First, pick your flavor
- Simulate a role when you’re missing expertise: “What would a behavioral scientist do here?” The persona advises you.
- Simulate a stakeholder when you’re missing a reaction: “How will a dairy farmer read this report?” The persona pushes back on you.
Step 1 — Create the persona
Don’t jump straight to “react to my report.” Start here:
I want you to simulate [a UK dairy farmer / a chief marketing officer for a small nonprofit].
Write a 3-paragraph description of this persona:
their background, daily pressures, what they care about, what they're
skeptical of, and how they talk.Step 2 — Refine the persona
You might know things the AI doesn’t. Tell it:
“Good, but farmers in our region are mostly tenant farmers, so land-use change is scarier for them. Also they’ve just been through a bad year for milk prices — factor that in.”
Or
“Great, but we don’t really do any marketing on Facebook, we mainly use LinkedIn and email.”
Every correction you make here improves everything the persona says later.
Step 3 — Save it
Turn the finished persona into something reusable: a Skill in Claude or ChatGPT, a Gem in Gemini. Paste the persona description into the custom instructions and add one line — “Always stay in this character unless I say ‘step out of character.’” Now you have a colleague you can go back to any time.
If your tool doesn’t have skills or gems, save the persona prompt in a document on your computer and paste it in whenever you need it.
Claude’s skills are covered in more detail in Set up Claude for your organization, including why an organization-wide skill should be reviewed before it is shared.
Step 4 — Use it
Stakeholder flavor — paste in your draft report, campaign page or letter:
Use the regional-journalist-persona skill.
Tell me: where you stopped reading, what made you roll your eyes,
what (if anything) landed, and what we could change to make it land better.
What would make you reshare this?Role flavor — bring your live problem:
Use the fundraising-consultant-persona skill.
We want to build on our recent aquaculture campaign and activate our donors
to fund the next step, which is a policy ask. How might we do that?Case study: Strat, Bryant Research’s strategic advisor
When VH AI Impact Director Richie was at Bryant Research, he built a senior strategic advisor persona that he informally called “Strat”, and used it to think through the company’s direction.
The instruction that made Strat useful: be no-nonsense, give honest feedback, and call out bad ideas. Richie and Strat would go back and forth — testing ideas, refining them, dropping the weak ones.
When Richie proposed training the whole Bryant team to build dashboards for animal protection organizations, Strat agreed that dashboards would be a useful service to offer. It disagreed that every team member should be taught to deliver them: several people weren’t technical, and it wouldn’t be a good use of their skills.
The most useful pushback was on ideas that were good but not great. Bryant Research already ran several services the animal movement found very valuable, and in a team with limited time, a new service has to be as good or better than the existing ones to earn a place on the roster.
How the VH team uses personas
In the Vegan Hacktivists AI services team, we built a change management consultant persona. One of our key jobs is advising animal protection organizations on how to increase AI adoption, and a big part of that is change management — the formal study of how organizations change and grow. Human change management consultants are expensive, so we simulated one. Ours reviews every AI action plan and every training session we run, checking that the advice we give leads to tangible improvements for the client.
The Amplify for Animals team uses a learning and development persona for the same reason. There is plenty of teaching and workshop experience on that team, but learning and development is an entire discipline with its own scholarship. The L&D persona reviews every Amplify slide deck and every piece of homework the team writes.
Choose someone AI can actually play
AI can only simulate people it has read about. As a rule of thumb:
- Works: professions and public archetypes (“a UK council procurement officer”, “a regional newspaper journalist”, “a fourth-generation beef farmer in Wales”), and public figures with a large written footprint.
- Doesn’t work: specific private individuals (“Sandra from the council”, your MP who never posts online). AI will invent a generic person and put your MP’s name on it — worse than useless, because it feels like insight.
If your real target is a specific person, simulate their role instead, then adjust for what you personally know about them.
Treat personas as inspiration
An AI simulation of a farmer is not a real farmer. Talking things through with a persona can change how you approach a piece of work, but treat what it says as a prompt for your own thinking rather than as evidence. Be especially wary of leaning on a persona for high-stakes decisions.
Have a real human validate your personas
If you find yourself using a persona all the time, that’s a good sign. The next step is to find a real human who fits it and ask them to help you refine it further. Having an actual senior marketing executive skim through your chief marketing officer persona can be a big help.
Advanced: build a council of personas
Why simulate a journalist at one major newspaper when you can simulate a journalist from every major newspaper? Why simulate one expert when you can simulate a whole board of them?
Once you have your first persona, create a few more — guardian-journalist-persona, daily-mirror-journalist-persona, times-journalist-persona, independent-journalist-persona. Then make one more skill, uk-media-personas, that runs all of them on the same piece of content.
What this is worth
An AI simulation of a consultant will rarely be as good as the real consultant. But a persona takes under a minute to set up, costs almost nothing, and only has to give you one idea for improving your work to be worth it. Compared with recruiting real experts or running stakeholder interviews, that is a very cheap way to make your work for animals land better.
Start with the stakeholder whose reaction you are least sure about, and build that one first.
