What Should Remain Human in AI-Supported Practice?
Session 2: Mapping a Real Workflow · 2 June 2026
What do we bring to our work as humans that a machine cannot replace?
This question guided the second AI4ST Practice Lab session, Mapping a Real Workflow, where a small group of researchers and practitioners came together to examine how AI is entering real sustainability transformations work.
Rather than asking how AI can be adopted more quickly, participants looked closely at their own everyday practices: data modelling, discourse analysis, qualitative research, facilitation, policy engagement, and creative writing. Together, they mapped workflows step by step, asking where AI might offer support, where it may introduce risk, and where human intelligence remains essential.
Starting from practice
The session was grounded in the realities of participants’ own work.
They reflected on the kinds of human judgment that remain essential across different areas of sustainability transformations work. In facilitation, this included the ability to read the room and respond to context in real time. In climate risk analysis and modelling, AI was seen as a possible thinking partner, but one that still requires human validation, local knowledge, and accountability. In discourse and qualitative analysis, participants noted that AI may support verification or pattern recognition, while interpretation remains rooted in human expertise. Creative writing raised another concern: AI often smooths language toward what is predictable and polished, risking the loss of voice, surprise, and meaning. In policy and stakeholder work, participants highlighted the relational and political dimensions of practice, where trust, power, and accountability cannot be automated.
Across these examples, a shared concern emerged: AI may be useful in parts of a workflow, but it cannot replace the human capacities that make transformation work situated, relational, creative, and accountable.
What is uniquely human?
Several themes surfaced through the session.
Contextual knowledge and judgment
Participants emphasised that sustainability transformations work depends on local knowledge, cultural understanding, and situated judgment. AI may process information quickly, but it cannot fully understand the norms, histories, relationships, and lived realities that shape responsible action in specific contexts.
Learning, struggle, and agency
The group discussed the risk of cognitive atrophy when AI is used to bypass difficult parts of the learning process. Struggling through a task is often how people build competence, confidence, and agency. If too much is automated too quickly, something important may be lost.
Creativity and the unexpected
In creative and reflective work, AI often moves toward what is grammatically clean, familiar, and statistically predictable. Human creativity, by contrast, often depends on the messy, surprising, irrational, and unexpected connections that carry meaning.
Relational trust and power
Participants also highlighted the deeply relational dimensions of transformation work. Building trust with communities, policy-makers, and partners requires presence, accountability, and an ability to navigate power dynamics. These are not technical tasks. They are human practices.
Collective emergence
The session itself also demonstrated something important: collective intelligence often emerges through conversation. Insights appeared through dialogue, disagreement, reflection, and the unpredictable movement of group thinking. This kind of emergence cannot be reduced to a machine-generated output.
Beyond AI adoption
One of the strongest insights from the session was that wise AI use does not mean using more AI. It means becoming more discerning about where AI belongs, where it does not, and what kinds of human capacities need to be protected and strengthened. Large language models are built on statistical prediction. Human ingenuity also depends on abductive reasoning, embodied experience, relational trust, collective emergence, and sometimes beautifully constructive errors.
For sustainability transformations work, the question is therefore not simply: What can AI do?
It is also: What kind of practice do we want to cultivate?
What comes next?
The AI4ST Practice Lab is a cumulative peer-learning journey. The next session will build on this workflow mapping exercise by looking more closely at the implications of AI-augmented practice.
Together, participants will explore what it takes to implement AI responsibly, what forms of trust and organisational support are needed, what safeguards may be necessary, and how these choices sit ethically and politically.
The next session will take place on 23 June 2026.
We warmly invite participants to join the next session and contribute to this growing community of practice.
