Gains:
- Ability to distinguish physical (acute/chronic) and transition risks and evaluate an institution together with its value chain in multiple warming scenarios
- Ability to match facility locations with climate hazard data and produce a qualitative risk level without having artificial intelligence match exact probability/amount
- Ability to honestly report the uncertainty of climate scenarios and avoid the risk of climate-washing
Climate change is not just an environmental problem, it is a financial and operational risk. A flood can shut down a factory, a carbon tax can collapse a business model, a drought can sever a supply chain. Institutions now have to measure these risks, test them with scenarios and report them. In this unit you will learn about the two main types of climate risk, scenario analysis, the role and limits of AI in modeling this uncertain future.
Frame first. TCFD (Task Force on Climate-related Financial Disclosures) is a world-standard framework that defines how climate risk will be disclosed; There are four pillars: governance, strategy, risk management, metrics-goals. (TCFD recommendations have now largely been transferred to ISSB standards, but their concepts remain fundamental.)
Two types of climate risks
Physical risk: risk arising from the physical effects of the climate. It is divided into two: acute (acute — sudden events such as floods, storms, heat waves) and chronic (chronic — slow, permanent changes such as sea level rise, average temperature increase, drought).
Transition risk: the risk arising from the transition to a low-carbon economy. Examples: carbon price/tax, regulatory change, technology shift (e.g. end of the internal combustion engine), change in consumer preference, risk of “stranded assets” (e.g. coal power plant becoming obsolete).
Tip: Physical and transition risk often move in opposite directions. If the world makes rapid mitigation, transition risk increases (sudden legislation), physical risk decreases; If he acts slowly, the opposite happens. That's why multiple scenarios are tested, not a single "future".
Scenario analysis: multiple futures
The future of climate risk is unpredictable, which is why scenario analysis is used: testing how soot will be affected under different warming paths (e.g. 1.5°C, 2°C, 3°C+). For scientific scenarios, IPCC's SSP (Shared Socioeconomic Pathways) and IEA's energy scenarios are taken as reference. The goal is not to “predict” the future, but to test resilience — “in which future will we break?”
AI is here; It is powerful at structuring scenario assumptions, processing different climate datasets, matching facility locations to physical hazard maps, and drafting risk scripts. But scenario selection, assumption realism and strategic interpretation belong to the expert.
Step by step: a climate risk assessment
1. Map assets and value chain. Which facility is where, which supplier is in which region?
2. Determine exposure to hazards. What physical hazards (flood, drought) is each location vulnerable to?
3. List the transition factors. Carbon price, legislation, technology, market.
4. Select scenarios. At least one low (1.5-2°C) and one high (3°C+) warming scenario.
5. Quantify impact. Potential financial/operational impact in each scenario (in range).
6. Assess and report durability. Weaknesses, adaptation measures, uncertainty.
three mini cases
Case 1 — Single scenario trap. One company told the AI to “calculate our climate risk” and it got a number over a single future. The consultant explained that this was misleading: in the low warming scenario the carbon tax (transition risk) was heavy, in the high warming scenario flooding (physical risk) was heavy. When the two scenarios were run separately, it became clear which future the company was vulnerable to. The odd number gave false confidence.
Case 2 — Fictitious possibility. An analyst asked YZ a facility's "What is the percent flood risk in 2050?" he asked; The AI gave an exact percentage. The expert noted that this number was not based on any model, but that the AI produced a number that “seemed reasonable.” The actual assessment required matching the location to a climate hazard map and hydrological model. The possibility of fabrication has been removed from the report.
Case 3 — Location matching. One group would assess the physical climate risk of 60 facilities. They had the AI write a Python workflow that matched facility coordinates to open climate hazard datasets; The code flagged 9 facilities with high exposure to floods and heatwaves. The team focused in-depth analysis on these 9 facilities. AI scanned; The expert made the final risk decision.
Weak prompt / Strong prompt
Weak prompt:
Evaluate my company's climate risk.
Why it is weak: No asset, location, scenario, risk type. The AI produces a generic script or a made-up “risk score.”
Powerful prompt:
Your role: climate risk advisor. With the TCFD framework, evaluate both physical (acute/chronic) and transit risk for the following list of facilities (location, sector). Use two scenarios: 1.5°C and 3°C. Even if it is definitely probable, it is FAKE; Give a qualitative level (low/medium/high) and add a note "what data should be quantified with". The result is the risk type × scenario table. Facilities: [list]
Four copyable templates
1) Risk type × scenario matrix:
Your role: climate risk expert. Evaluate the following physical (acute/chronic) and transition risks for the institution in two scenarios (low/high warming). Give a qualitative level and brief justification in each cell. Don't make exact numbers; specify what data is needed in the quantification. Institution information: [here]
2) Location-hazard matching (by code):
Write a Python workflow framework that maps the following facility coordinates to open climate hazard datasets (e.g. flood, heat wave). Flag facilities with high exposure. Specify the data source; do not fit the score, put a note "must be validated against the data set". Facilities: [here]
3) Transition risk factor screening:
List transition risks for the following business model: carbon price, regulatory, technology shift, market/consumer, sunk asset, reputation. For each, specify exposure and possible time horizon (short/medium/long). Fitting exact amount. Business model: [here]
4) Draft resilience and adaptation measures:
Suggest possible adaptation/resilience measures against the following identified climate risks. For each measure: which risk it reduces, in which scenario it will work, rough level of effort. Don't use the hyperbolic phrase "zeroes the risk". Risks: [here]
Common mistakes
- Evaluating with a single scenario. At least one low and one high warming scenario is required.
- Mixing physical and transition risk. The two are analyzed separately and may be in opposite directions.
- Making the AI fit exact probability/amount. Quantification requires real models and data.
- Bypassing the value chain. The risk often lies with suppliers and customers.
- Hiding uncertainty. Climate scenarios are inherently uncertain; report this.
Caution: Climate risk disclosures influence investor decisions, and false/exaggerated disclosure is the financial version of greenwashing — “climate-washing.” Underestimating the risk or presenting it with false certainty created by AI both create legal liability.
In summary
Climate risk; It requires assessing physical (acute/chronic) and transition risks along with the value chain in multiple scenarios. AI; It is a powerful accelerator for location-hazard matching, factor screening, and drafting. But scenario selection, quantification based on the real model, honest reporting of uncertainty and strategic interpretation belong to the expert. The aim is not to know the future, but to test endurance.
Application task
List three or four facilities of an institution with their locations and indicate their industry. 1. create a matrix with the template that tells the AI to evaluate physical and transition risks in two scenarios; Verify that the AI is not making up exact probability but giving qualitative level. Then, with template 3, scan your business model's transition risks and mark the three risks with the shortest horizons.
checklist
- [ ] I evaluated physical (acute/chronic) and transition risks separately.
- [ ] I used at least one low and one high heating scenario.
- [ ] I included the value chain (supplier/customer) in the analysis.
- [ ] I didn't make the AI fit the exact probability/amount; I specified data for quantification.
- [ ] I reported the uncertainty clearly.
- [ ] I have taken the risk of climate-washing the description (neither slight nor fake for sure).