WWF International: putting AI to work for a global media team

Case study · AI training

WWF International’s media relations team wanted to use AI properly, inside strict governance rules. We built a three-session Microsoft Copilot programme around their real work, from the daily media digest to crisis triage. At the end, their media relations manager introduced us to WWF’s Head of Creative and Content.

• 3 live sessions
• 6 deliveries across time zones
• 8 participants
• 4 countries

• Client WWF International
• Sector Conservation charity
• Training AI for PR and comms, Microsoft Copilot, prompt pack
• When July to September 2026
• Audience International media relations and issues team
• Format Three 90-minute Teams sessions, each run twice

What did WWF need?

WWF International’s media and issues management team works across Brazil, Singapore, the Netherlands and the UK. They were already experimenting with AI, but WWF’s internal AI training focused on conservation, not communications.

The rules were tight. Microsoft Copilot was the only approved tool, new AI uses need central sign-off, and the environmental cost of AI has to be weighed. WWF found us through AI search.

• Business objective: save the team time and protect WWF’s reputation as AI changes how news is made.
• Training objective: give every participant the confidence to use Copilot in day-to-day media work, with clear use cases that fit WWF’s governance. Success measures agreed at kick-off: more confidence, practical adoption, clear use cases, and feedback from participants and leadership.

What was the insight?

Our pre-training survey, answered by seven of the eight participants, showed a team “largely past the awareness stage”. Six already used Copilot Chat. Their biggest worry, by far, was accuracy and hallucinations.

So the programme didn’t sell AI. It taught them to check it. We built SAFE, a simple test for every AI output: Source, Accuracy, Fit, Escalate.

What did the training cover?

• Session 1: AI, journalism and everyday Copilot. How newsrooms use AI, what that means for pitching, AEO and GEO, and hands-on Copilot for the daily digest and media writing in WWF’s tone.
• Session 2: responsible AI, media relations and issues. Copilot app by app, media lists and tailored pitches, issues triage and a live crisis simulation built on a fake post about WWF donations.
• Session 3: agents. When to use a prompt, a notebook, a scheduled prompt or an agent, then building and testing an onboarding agent end to end.
• Built to fit the team. Each session ran twice, morning and afternoon, so every time zone could join. Session 1 feedback reshaped Session 2. Everyone left with a bespoke prompt pack, the decks and the recordings.

What did it achieve?

Results by stage of the AMEC evaluation framework, with the source for each figure. Outcome measures are due at the follow-up in November 2026.

What did WWF get from it?

A team working with AI inside WWF’s rules rather than around them, with a shared checking habit and prompts built for its own tasks.

The clearest sign came at the end. WWF’s media relations manager introduced us to the Head of Creative and Content, who had “heard wind of our impressive training” and wants to explore support for his team.

“It’s been so excellent and you’ve been so flexible.”

What did we learn?

• Start from what they already do. The survey showed the team was past the basics, so we skipped AI 101 and went straight to real tasks.
• Treat governance as a feature. Building around Copilot-only rules made the advice usable on day one.
• Let feedback shape the next session. Requests for more practical application after Session 1 changed Session 2.
• What we would do differently. Agree the follow-up date at the start. The check-in slipped behind WWF’s busiest season, so outcome data is still to come.

Want your comms team using Copilot safely and well? Talk to us about AI training.