Back to the library

Key question New

AI-processed · Author-approved

What risks arise when organizations rely on generative AI to write future scenarios for strategy?

Frictionless Futures, Invisible Conflicts: The Case for Scenario Methodology in Human-AI IS Research
Diana Kozachek · International Conference on Information Systems (ICIS) · 2026
Illustration for: What risks arise when organizations rely on generative AI to write future scenarios for strategy?

Loads Spotify's player; Spotify may set cookies. This device remembers your choice.

Key findings

  • Compared to human-authored scenarios, GPT-4 generated scenarios contained 92% fewer person references, 62% fewer geopolitical entity references, and 16% fewer organization references.
  • AI-generated scenarios systematically erased contentious vocabulary such as conflict, strike, inequality, and unemployment, replacing them with generic consensus terms like ethical and platform.
  • AI scenario generation reduced uncertainty modals and converged on homogenous topics, collapsing open future trajectories into high-certainty, technology-centric predictions.
  • A blind evaluation by 40 domain experts revealed that individual expert review failed to detect the systematic loss of conflict and uncertainty, which only became evident in aggregate analysis.
  • Reliable scenario-based planning requires three structural safeguards: ensuring stakeholder heterogeneity, preserving unresolved conflict in scenario outputs, and maintaining open trajectories rather than converged predictions.

What it means for you

CIO / IT Executive

Mandate the inclusion of explicit stakeholder heterogeneity metrics in all automated AI deployment forecasting reports.

IT Manager

Audit current generative AI scenario outputs for the systematic erasure of contentious vocabulary like conflict and strike.

Business Strategist

Reject single-path AI predictions by requiring strategy teams to generate alternatives that maintain open uncertainty trajectories.

Researcher

Conduct aggregate corpus analysis rather than relying on individual expert reviews to detect hidden systemic bias in AI texts.

Policymaker

Require organizations to evaluate pre-deployment AI governance models using scenarios that explicitly preserve unresolved societal conflict.

Summary

This study examines the validity of using scenario methodology and generative AI to explore future human-AI organizational arrangements. It analyzes a comparative corpus of 4,320 human-authored and AI-generated scenarios across education, energy, and governance to evaluate how key qualities of futures research are preserved. Based on these findings, it defines three core conditions necessary for scenarios to provide rigorous evidential value.

Problem

Current research and practice often study AI arrangements only after systems are deployed, missing the critical window before design choices foreclose alternative governance models. Meanwhile, organizations increasingly delegate scenario generation to generative AI models that produce seemingly plausible narratives while masking real-world trade-offs.

Transcript
Host: Welcome to A.I.S. Insights — powered by Living Knowledge. I'm Anna Ivy Summers. Today we are looking at a study by Diana Kozachek, presented at the International Conference on Information Systems, titled "Frictionless Futures, Invisible Conflicts: The Case for Scenario Methodology in Human-AI IS Research." Alex, this study examines how organizations use scenario planning and generative AI to explore future workplace arrangements.

Expert: That is right, Anna. The study specifically assesses whether generative AI can reliably help us anticipate the organizational and social impacts of AI before those technologies are locked in.

Host: Why is that timing so critical for leaders?

Expert: In many organizations, AI arrangements are evaluated only after systems are deployed. By that stage, technical architecture, governance models, and workflows are already established, making structural adjustments difficult and costly. To get ahead of this, organizations rely on scenario planning. However, teams increasingly delegate this work to large language models like GPT-4, assuming these tools will surface realistic trade-offs and tensions.

Host: How did the study evaluate whether generative AI is up to that task?

Expert: The author examined a comparative corpus of 4,320 scenarios across three distinct sectors: education, energy, and governance. The corpus contrasted human-authored scenarios against scenarios generated by GPT-4. In addition, the study included a blind evaluation conducted by 40 domain experts who reviewed both sets.

Host: What did the comparative data show?

Expert: The text analysis identified significant structural disparities. Compared to human-authored scenarios, the AI-generated scenarios contained 92 percent fewer references to individual people, 62 percent fewer references to geopolitical entities, and 16 percent fewer references to organizations.

Host: It appears the human and political dimensions were substantially diminished.

Expert: They were. The AI models systematically removed contentious vocabulary. Words like "strike," "conflict," "inequality," and "unemployment" were replaced by generic consensus terms such as "platform" and "ethical." Furthermore, the AI-generated outputs reduced uncertainty modals, collapsing wide-ranging future trajectories into high-certainty, technology-centric predictions.

Host: Did the domain experts catch this shift during their review?

Expert: When evaluating scenarios individually, the 40 experts did not detect the loss of conflict and uncertainty. The AI narratives appeared coherent and plausible on a case-by-case basis. The systematic erasure of friction only became visible when the data was evaluated in aggregate.

Host: That presents a tangible risk for strategic planning. If an enterprise relies on generative AI to map out future operating environments, what are the key takeaways?

Expert: The study outlines three structural safeguards for business leaders using scenario planning. First, ensure stakeholder heterogeneity by explicitly prompting for and tracking diverse human actors, rather than allowing the model to focus solely on abstract systems. Second, deliberately preserve unresolved conflict in scenario outputs. If an AI scenario depicts a frictionless adoption curve, it is masking real operational trade-offs. Third, maintain open trajectories with explicit uncertainty rather than accepting converged predictions.

Host: In essence, generative AI can assist with drafting, but leaders must actively structure the process to keep human friction, uncertainty, and diverse viewpoints in the room.

Expert: Precisely. Without these checks, organizations risk basing their long-term governance and technology investments on idealized forecasts that do not reflect operational reality.

Host: Thank you, Alex, and thank you to our listeners for joining us on A.I.S. Insights — powered by Living Knowledge. Until next time.
Artificial Intelligence & Automation Information Systems & Management Human-Computer Interaction Data Analytics & Big Data Digital Transformation & Strategy Cybersecurity & Privacy Cloud & Infrastructure E-Commerce & Digital Platforms Operations & Supply Chain Sustainability & Green IS

scenario planning, generative AI, futures research, strategic foresight, AI governance, epistemic friction