Methodology
300 personas, 300 scenarios,
20 digital twins, 5 composite digital twins
The scenario planning workshop utilized diverse participant groups to generate and evaluate scenarios, leading to four distinct sets of perspectives:
Personas (The Individuals)
This set comprised 1,000 unique descriptions of people actively engaged in transformative innovation across agriculture, climate, biodiversity, and AI products. These individuals were divided into categories based on their professional focus, including:
People directly engaged in innovation (40%).
Sophisticated generalists in policy, product, management, or data roles (20%).
Individuals in the digital ‘information industry’ (10%).
Adjacent stakeholders like investors, academics, and policy developers (10%).
People with particularly imaginative perspectives, such as artists and students (10%).
Digital Twins (Real People Representations)
Descriptions of 20 specific real people (e.g., Yuval Noah Harari, Mustafa Suleyman, Mark Joseph Carney, Gabe Brown) were created, synthesizing their views, expertise, styles, and motivations based on thorough research. These representations explicitly excluded identifying information.
Composite Digital Twins (The Workshop Actors)
This set consisted of 20 synthetic 'composite digital twins' (e.g., The Existential Technocrat, The Regulatory Realist, The Food System Architect). Each composite blended characteristics from three different actual digital twins to ensure a coherent and plausible blend of real-world expertise and perspectives for workshop simulations.
The Judging Panel (The Evaluators)
A panel of five judges (drawn from the Composite Digital Twins) was established to independently evaluate the 1,000 short scenarios. They scored each scenario on a scale of 1 to 7 using four criteria: Imagination & creativity, Possibility, Probability vs. impact, and Robustness.
300 scenarios, 3 simulated workshop discussions, analysis and report generation
The process then focused on large-scale idea generation and systematic refinement of foundational drivers. The one thousand (300) Personas created short scenarios (200-300 words) detailing a fundamentally transformed AI-mediated future for agriculture.
Each scenario specified a specific ‘end state’ and two to four causal ‘driving forces’ related to AI functionality. A judging panel comprised of five Composite Digital Twins independently evaluated these scenarios. This assessment identified 30 'driving forces'—the most common and impactful external factors and trends—which became the structural foundation for the subsequent workshop phases.
The project then transitioned into sequential, simulated workshops for deep analysis. The first phase (Prompt 6) involved ten groups analyzing assigned driving forces, assessing their implications, underlying factors, and rating their unpredictability and impact on a 1-to-7 scale.
This was followed by a second phase (Prompt 7) where groups envisioned and explored specific ‘End States’ resulting from those drivers, focusing on radical future characteristics.
The transcripts were then subjected to an ‘affinity mapping’ analysis (Prompt 8) to synthesize the collective findings into 10 coherent ‘proto scenarios’—macro-scale end states for the agricultural system.
Finally, a large group analyzed the specific dynamics (accelerators, counterforces) and detailed the critical implications and possible responses for each of the 10 proto scenarios (Prompt 9 and 10), completing the structural foresight analysis
Notes on the methodology
and limitations
In the process, a few shortcomings of the approach are worth highlighting.
The original aim was to generate one thousand separate AI personas from different predefined categories. The personas themselves appeared plausible, but after 300 started to repeat,. The same pattern appeared when generating scenarios where after a while, the same themes and text kept surfacing. Personas and scenarios were therefore limited to 300 and any repeats excluded from the workshop participant pool.
When simulating workshop discussions—despite being provided with personas, backgrounds, and scenarios that the personas had generated—the conversation was not natural, lacked the depth and breadth that you would expect to find in a one-hour session, and most importantly lacked any personal or anecdotal lived experience.
The digital twins (full list below) used to create 'composite' digital twins may have influenced the final scenarios too strongly. The were semi-arbitarily picked to cover a range of perspectives - but still the selection (eg mostly European/American, 'progressive') may have biased the outcomes. Nevertheless, it would be impossible/difficult to bring groups like this together, which is one of the strengths of an AI generated approach.
Yuval Noah Harari – Historian and public intellectual who has written and spoken about AI’s potential to hack and reshape human society.
Mustafa Suleyman – Co-founder of DeepMind and CEO of Inflection AI (now Microsoft AI), Suleyman is both an AI creator and thought leader on AI’s societal risks.
Gabe Brown – one of the pioneers of the current soil health movement which focuses on the regeneration of natural resources.
Ivo Degn – the co-founder and Managing Director of Climate Farmers.
Mark Carney – Politician and economist, current prime minister of Canada.
Indy Johar - an architect, co-founder of 00 (project00.cc) and most recently Dark Matter Labs.
Dieter Helm - Professor of Economic Policy at the University of Oxford and Fellow in Economics at New College, Oxford
Hans Stegeman - Economist & Executive Leader at Triodos Bank.
Henry Dimbleby - Co-founder of Bramble Partners. Formerly co-founder of Leon.
John Cutler - Writer on cross-functional product development. Ex Amplitude
Patrick Holden - Executive Director of the Sustainable Food Alliance.
Thomas Gent - Farmer and Founder of the award winning Gentle Farming brand. Key player in developing the UK and European carbon markets.