Co-designing AI policies: why the process matters as much as the output

The illusion of quick certainty

Artificial intelligence is everywhere, and the noise can feel overwhelming. Leaders are often told that these tools will change everything, presenting massive opportunities alongside complex social and ethical challenges. In response, many organisations rush to create a top-down policy to regain control.

When teams face uncertainty, they naturally look for black-and-white answers. AI tools are fantastic at providing confident responses, but they are statistical models rather than creative thinkers. Relying on quick answers can cause us to skip the messy, essential conversations that build trust and real understanding.

What makes an organisation effective will also make it good at exploring AI. Staying user-centred, collaborative and curious is far more valuable than becoming a technical expert overnight.

The value is in the process, not the final output

When we developed our own approach to AI at William Joseph, we knew a top-down policy would not fit our culture. Instead, we brought our team together in workshops to co-create our framework (wonderfully facilitated by Matt, Ray and Ross). We realised that the primary benefit was in the shared understanding created during the discussions, not just the document we produced at the end. 

A co-creation process gives everyone a space to voice their ethical worries, emotional reactions and internal conflicts. People might like the speed of a tool while hating the origin of its training data. Opening up room for these nuanced feelings removes anxiety and prevents people from feeling isolated.

If you are worried about these tools, your audiences, clients and beneficiaries likely are too. Involving your whole team - and your board of trustees - ensures your policy reflects real, lived perspectives rather than abstract assumptions.

Holding space for wicked questions

To run an effective co-design session, you need to hold space for opposing yet complementary ideas. Ray used the “wicked questions” approach from Liberating Structures to explore complex problems that do not have a single right answer. These questions are framed to hold two truths at once, such as “How can we innovate using AI while protecting creative integrity?”

During our co-creation sessions, we explored several wicked questions that helped map out our grey areas. Examples included:

• How can we use AI to make our work more efficient without compromising on quality?
• How can we maintain our Equity by Design principles while using tools that reproduce harmful biases?
• How can I use AI to help me explore unfamiliar topics while still validating that its output is accurate?
• How can we maintain our people-first approach while simultaneously embracing new automation tools?

Exploring these questions together allows teams to navigate ambiguity safely. It shifts the mindset from trying to eliminate risk to intentionally managing it as a group.

Framing principles around shared values

Out of our co-creation workshops, we established guiding principles grounded in our core mission. An AI policy should never be about the tools alone; it must be guided by your values.

Our approach rests on seven core commitments:

• Human-centred and expert-led: specialist knowledge and empathy remain our greatest assets, so AI outputs must always be reviewed by a human expert.
• Transparency and accountability: trust is our most valuable currency, so we are open about which tools we use and remain fully accountable for our work.
• Curiosity and continuous learning: we create time and space for safe, playful experimentation where failures are shared openly without judgement.
• Client-focused and responsible: our exploration is driven by client needs, and we never recommend solutions we have not tested and understood.
• Mindful and critical use: we remain specific about individual tools and question all outputs to verify their accuracy.
• Supportive and collaborative culture: we support each other without pressuring colleagues to adopt tools they feel uncomfortable with.
• Inclusion: we seek to realise the democratising power of new tools to reduce barriers, while actively challenging internet biases and harmful stereotypes.

Set up a community of practice around your policy

An AI policy created in isolation becomes obsolete almost immediately. Because technology and user needs evolve rapidly, your policy must be treated as a living document that is reviewed regularly with your team.

Creating communities of practice across your organisation allows people to continue sharing learnings, testing use cases and refining standards together. This continuous feedback loop ensures that your governance remains supportive rather than restrictive.

By focusing on process, collaboration and psychological safety, you can build an AI approach that strengthens your team culture while keeping your mission at the heart of every decision.