This post shares practical reflections on how AI tools can support those working in contexts where collaboration, co-design, and facilitation are important. Drawing from experience, I outline four ways these tools have become quiet thought partners in my work – and reflect on where their strengths end and human judgement, relationships, and systems thinking still matter most.
In my own practice, tools like GPT, Claude and Perplexity have become quiet companions in the background. I’ve been using AI in this space since it first became more widely available in early 2023, not to outsource my thinking, but to support it. These tools have helped me with framing, exploring, writing, and preparing across a wide range of projects.
In practice, I’ve found AI tools to be especially helpful when working through ideas – whether I’m drafting a blog post, curating material for the Learning for Sustainability site, designing a workshop, or synthesising interviews and field notes. They help me surface insights, test interpretations, and gain clarity more quickly.
I was struck recently by a post from Michelle Wiles reflecting on AI’s role in consulting. She writes:
“AI will be used as a thought partner to speed up research and get to a plan faster… Persuading teams, building relationships – that’s not work an AI can do (yet).”
While her context is corporate strategy, the insight carries over. For those of us working in participatory, systems-facing settings, much of the work is relational. It involves framing, convening, and supporting collaborative thinking and action. AI doesn’t replace that, but it can help us prepare more thoughtfully for it.
Here are four ways I’ve been using these tools in my own practice:
1. Exploring new research and perspectives
As someone who curates open-access resources for the Learning for Sustainability site, I’m regularly reading across disciplines. But when I need to explore a new topic, trace emerging framings, or look for fresh insights, GPT and Perplexity have become helpful companions.
With the right prompts, they can surface recent academic articles, policy papers, and blogs that might not turn up in a standard search. They can also suggest related concepts, authors or bodies of work that I may not have thought to look for, which can help me move beyond my initial framing of a topic. These tools can still invent, misidentify or misrepresent sources, so I check and read the sources before drawing on them.
It’s not a shortcut for reading. The value is more in broadening the field, helping me see where else to look, and sometimes showing me that the question I started with was too narrow.
2. Supporting systems framing and complexity
Framing, not fixing. In systems work, I often find myself dealing with issues that don’t have tidy answers. Whether I’m preparing a theory of change, shaping a facilitation run-sheet, synthesising interviews and field notes, or developing reflective prompts for group work – these are all acts of framing, part of the wider systems thinking and systemic design practice I describe here.
AI can surface possible framings, contrast assumptions and reveal tensions I might otherwise miss. I still need to decide what is useful, what fits the context, and what I am prepared to stand behind. Different framings can make different relationships, trade-offs or blind spots more visible.
Sometimes I use it to trial different starting points for a session, or to test how a set of themes might cluster. Other times it helps me rehearse arguments or explore ways to present a tricky issue. The back-and-forth can help me see where a framing is too narrow, where an assumption needs questioning, or where another perspective might change the way the issue is understood.
3. Making writing less lonely
When writing, I often start with scattered notes and phrases. I may use AI to explore possible structures or ways of expressing an idea, reject much of what comes back, follow a connection I had not thought of, and sometimes rethink the argument altogether. In that kind of exchange, it is not always possible, or particularly useful, to draw a clean line between where one contribution ends and another begins. Once the direction becomes clearer, I work closely through the structure, paragraphs and sentences, although that can still send me back to reconsider the thinking.
For me, the value lies partly in that back-and-forth. The important question is whether I am still questioning, making connections, changing my mind and working out what I actually think. By the time I publish, I need to be satisfied that the piece says what I mean, that I have worked through the argument rather than simply accepted it, and that I am prepared to take responsibility for it.
This is not entirely different from writing with co-authors. In a good collaborative process, ideas develop through discussion, challenge, rewriting and response. After several rounds, it can be difficult to say exactly whose idea a particular point was, or how it evolved. What matters is that the thinking has been actively worked through, rather than simply accepted because someone, or something, proposed it.
This way of thinking about AI as part of a collaborative process is also beginning to appear in research on human–AI writing and “thought partnerships”, including work on how interaction with AI can support ideation, sensemaking and the development of ideas, while also raising questions about agency and over-reliance.
4. Testing how things land
Another way I use these tools is to explore how a piece of writing might be interpreted by different audiences. I might ask: How might this come across to a biophysical scientist? A social researcher? A policymaker? A practitioner working on the ground? I would not take these responses as reliable predictions of how actual people will respond, but rather as one way of stretching perspective and looking for possible blind spots.
This has become a regular part of how I write and review, especially when preparing public-facing material or strategic documents. It can help me clarify tone, consider posible misreadings, and think about whether the language could be more more inclusive. But it can’t stand in for lived experience or cultural ways of knowing, and it’s important to stay mindful of that, especially when working in equity-focused or cross-cultural spaces. This kind of reflective adaptation is key to how I think about co-design in complex settings.
Working with AI tools, I’ve found it important to think carefully about the ethical implications, especially around data sensitivity, privacy, and being transparent about when and how these tools are used. I’m also mindful of concerns around AI-generated content and source attribution. This includes being alert to the risk of unknowingly drawing on material that hasn’t been shared with permission. In work that values transparency and shared knowledge, these are important ethical boundaries to navigate.
Working in complementary ways
The examples above highlight how AI tools can support real work without replacing it. In practice, I’ve found it helpful to think about where these tools complement – rather than compete with – human judgment, facilitation, and collaboration. The table below sets out some of those distinctions, based on how I use AI in day-to-day practice.
Table: How AI can contribute to collaborative research and facilitation
| Task/Role | How AI may contribute | What still needs human judgement |
| Literature search & synthesis | Rapid scanning, surfacing new sources | Critical reading, contextual judgement |
| Qualitative analysis & synthesis | Summarising interviews, clustering themes | Meaning-making, context, ethical interpretation |
| Framing & systems thinking | Suggesting framings, testing assumptions | Coherence, relevance, group-based interpretation |
| Drafting & rewriting | Contributing potential structures, arguments, connections and wording | Questioning, selection, revision, voice, meaning and responsibility |
| Testing audience response | Simulating reactions, checking tone | Knowing what matters, inclusive framing, cultural nuance |
| Facilitation & relationship work | Contributing prompts, synthesis and other inputs within deliberately designed group processes | Trust, empathy, power dynamics, situational judgment |
What remains ours
Looking across that table, AI can contribute to analysis, drafting and exploration, but it does not take on the human responsibilities that sit at the heart of participatory and systems-facing work.
AI does not itself hold relationships, understand what is at stake for people, or take responsibility for navigating power and consequence. Those remain human responsibilities. They require trust, timing, emotional intelligence, and a deep understanding of context.
In collaborative research, evaluation, and systems change work, success isn’t just about generating insight. It’s about connecting the dots – spotting patterns, surfacing tensions, and helping people find clarity in complexity. It’s also about influence: shaping decisions, earning buy-in, and building shared momentum around ideas that matter. These are themes I’ve reflected on in more depth here.
At the core, what matters most is not the tool itself, but how we use it – and who we use it with. For me, these technologies work best when they support rather than displace the thinking, judgement, relationships and learning that the work depends on.
More information on the use of AI can be found through the LfS generative AI landing page. This includes links to the related resource pages Using AI in research and practice and AI in context: the wider picture. A few related posts from that collection include:
- AI prompts for shared thinking: a light framework for purposeful prompting
- Working with AI in the room: authorship, responsibility, and collective judgement
- Seeing the wider ethical picture around AI development and use
- AI in place-based practice: what is shifting
[Updated August 2026 to reflect how AI tools have evolved, and how my thinking and practice have developed, particularly around research, systems framing and writing.]
[* Photo by:ThisIsEngineering | Pexels]
