Scenario Analysis in the AI Era: Redefining Human Involvement - Quick Summary

nathanalbinagorta
Jun 27
4 min read

Here is a summary of the paper "Scenario Analysis in the AI Era: Redefining Human Involvement" published in May 2025 by George A. Shinkle, Chirag Gujarati and Patrick Sharry. George mentioned this paper in the last AGSM Business School professional development. Loved the fact that AI is treated here not just in the context of operations, but a key piece in the strategy formulation field.

Shinkle, Gujarati, and Sharry explore how the strategic management practice of scenario analysis can be transformed and enriched by generative AI technologies. The authors focus on two key questions:

How can humans be effectively integrated (“humans-in-the-loop”) when using generative AI to develop plausible future scenarios?
Once scenarios are generated, how should organizations decide on prudent responses?

To answer these, they develop two managerial tools: the Scenario Analysis Guidance Tool and the Scenario Response Spectrum Framework. These are designed to help organisations not only generate relevant scenarios using AI but also take informed, structured action in response to them.

1. The Changing Role of Scenario Analysis

The paper begins by contextualising the value of scenario analysis in turbulent environments. Traditionally, organizations have used scenario planning to help leaders consider uncertain futures and test strategic flexibility. However, recent advances in generative AI have disrupted this field by enabling the rapid creation of plausible future scenarios at virtually no cost.

While AI brings efficiency and breadth to scenario generation, it lacks sensitivity to organisational context, ethics, and strategic nuance. This creates a new imperative: combining AI’s speed and data-processing with human intuition and contextual insight.

2. The Scenario Analysis Guidance Tool

This tool is developed to help managers thoughtfully use AI in scenario generation. It is structured around three phases: prompting, testing, and tuning.

Prompting

The first phase involves instructing AI to generate scenarios. Three styles of prompts are proposed:

One-shot prompts: A single, broad request to produce multiple scenarios.
Chain-of-thought prompts: A stepwise method involving identification of trends and uncertainties first, then combining them into scenarios.
Self-critique iteration: An AI-generated critique of its own output followed by refined scenarios.

Testing

Here, the authors emphasise evaluating AI-generated scenarios for relevance, plausibility, and managerial interpretability. Questions include:

Do the scenarios cover critical uncertainties?
Are they plausible representations of future states?
Can managers understand and relate them to current operations?

Tuning

The final step ensures scenarios are “fit for purpose.” Scenarios are refined for appropriate scope (short- vs long-term), complexity, and emotional tone (too optimistic or too threatening). This step ensures organizational buy-in and contextual fit.

The authors stress that customisation is necessary, as AI outputs vary and organisations differ in purpose, culture, and strategic needs. Generative AI should be seen as a drafting partner, not a definitive authority.

3. The Scenario Response Spectrum Framework

After generating scenarios, organisations face the crucial task of deciding how to respond. The authors criticise existing literature for assuming that scenario exposure alone leads to better decision-making. Instead, they propose a structured, three-phase framework to assess and prioritize responses:

Phase 1: Impact Assessment

Scenarios are evaluated across four dimensions (scored 1 to 5):

Probability: Likelihood of occurrence
Repercussion: Negative impact on current business
Urgency: Time pressure to respond effectively
Strategic Disruption: Likely impact on future strategy

These scores are totaled to assess each scenario’s impact.

Phase 2: Organisational Risk Tolerance

This phase accounts for the organisation’s risk appetite (willingness to take risk) and risk capacity (ability to absorb risk). Combined into a total “risk tolerance” score, this moderates how impact scores translate into action.

Risk tolerance is influenced by leadership style, industry norms, stakeholder pressure, and organizational resources.

Phase 3: Response Spectrum Evaluation

Scenarios are categorised into five response types:

Priority Action: Take immediate action.
Timely Action: Plan action in the near future.
Safeguard: Develop contingency plans.
Monitor: Watch for early warning signs.
Ignore: Take no action (for now).

The evaluation balances scenario impact with organisational risk tolerance, enabling more calibrated, realistic decisions. Notably, scenarios with high impact might still receive moderate responses if the organisation has low risk tolerance, or vice versa.

4. Application and Practitioner Insights

The authors developed and refined these tools through MBA teaching and consulting engagements. Key insights include:

AI is a strong brainstorming partner, capable of quickly producing diverse scenario sets that stimulate group discussion.
Tuning is vital: Scenario tone, complexity, and scope must match the audience and purpose.
Scenarios should prompt strategy, not just discussion: The framework enables this by translating scenarios into concrete action categories.
Scenario overload can be managed: By generating more than the traditional three to four scenarios (e.g., 10), then narrowing them down for deeper exploration, AI-generated breadth becomes a strength.

5. Broader Implications in the AI Era

The authors foresee three major changes to scenario planning:

Continuous Scenario Development: With AI’s speed, organisations can move from occasional scenario planning to ongoing strategic foresight (monthly, even real-time).
Expanded Scenario Sets: AI enables the use of morphological matrix approaches to explore combinations of uncertainties at scale, leading to more robust preparedness.
Scenario Planning as a Core Strategic Practice: As uncertainty grows in a digital world, scenario analysis becomes more essential, despite or because of AI automation.

They summarise these into three conjectures:

Scenario development will align with industry change cadence, leading to continuous strategic planning.
Organisations will use AI to generate more scenarios, enhancing comprehensiveness.
Scenario analysis will grow in strategic importance in the AI-augmented future.