AI scored 50 banks on climate adaptation - what we learned
Louie from Climate Proof had a problem.
He wanted to know which of the world's 50 largest banks were genuinely preparing for physical climate risk. Not just talking about it, but actually doing the work.
The challenge: finding and assessing evidence scattered across tens of thousands of pages of annual reports and disclosures.
He partnered with us at ClimateAligned to build a custom analysis workflow in brief.green, designed for qualitative research that needs both scale and specificity.
I sat down with him to talk about what he learned.
“When you look at what banks are actually disclosing—it's not clear who's doing the work.”
Climate adaptation is finally getting traction in banking with regulatory influence and industry groups like UNEP FI publishing adaptation guidance.
But Louie kept noticing a gap: "There's all this talk at the policy level. But when you look at what banks are actually disclosing, what they're willing to put in their annual reports, it's not clear who's doing the work."
He wanted evidence. Not promises, not press releases—actual proof of who was integrating adaptation into core operations.
The problem: that evidence was scattered across 50 banks, 15 different indicators, and hundreds of documents totalling tens of thousands of pages.
"The thing I appreciated most was that I could see the reasoning”
Louie designed a framework with 15 qualitative indicators and 124 subcomponents, including physical risk assessment maturity and awareness of national adaptation policies.
He worked with our team to encode those indicators into a brief.green workflow. The AI read each bank's disclosures, extracted evidence, scored each subcomponent, and cited back to the source documents.
“The thing I appreciated most,” he said, “was that I could see the reasoning. Every score was tied to a specific quote from a specific document. If I disagreed, I could trace it back and understand why the AI scored it that way.”
"Some days I had to run [the workflow] multiple times"
“For most banks, the AI nailed the structure on the first run—all 124 subcomponents, properly formatted. But for some, I had to run the assessment a few times to get consistent output.”
He paused. “That's not a failure, though. It's just the reality of working with LLMs. They're probabilistic. Even with tight guardrails, there's variance.”
His suggestion: run each assessment five times automatically and show users the range. “Give people an error margin. Acknowledge the subjectivity. That would build even more trust.”
I asked him how this compared to doing it manually.
“If I'd hired a team of interns, we'd have the same problem, maybe worse. You rate the same company differently depending on whether you've had coffee that morning.”
“At least with the AI, I can bake my assumptions into the workflow. Everyone can see the rules I set. That's more transparent than pretending human judgment is consistent.”
“[The data] shows that slow, persistent work by regulators and voluntary groups does eventually move the needle”
Banks are better at physical risk assessment than expected. Even laggards in the US, Canada, and Australia showed high maturity. Why? A decade of TCFD/IFRS influence and regulatory stress testing.
“It shows that slow, persistent work by regulators and voluntary groups does eventually move the needle,” Louie said.
But huge gaps remain. Most banks aren't engaging with national adaptation plans. They mention "climate policy" but don't distinguish mitigation from adaptation.
“This year at the policy level, it's all been about adaptation. But that's not flowing through to bank disclosures. Only a few banks—like Credit Agricole, which explicitly referenced France's adaptation plan—are making that distinction.”
More banks than expected are engaging clients on resilience. Some offer green mortgages that promote energy resilience, others have helped launch green bonds that finance climate-proofing initiatives.
Standard Chartered topped the list with an average indicator score of 88%. This may in part reflect the bank's exposure to coastal assets in Asia, which are at high risk from climate shocks. The bank has also actively promoted adaptation financing in recent years, making it a rare outlier in the sector. European banks like BNP Paribas and Intesa Sanpaolo also performed well. So did National Australia Bank, facing massive drought, fire, and storm risks.
The pattern is that banks with the most exposed portfolios understand both the risk and the revenue opportunity in adaptation finance.
Read the complete findings here
What this means for your work
If you're doing sustainability benchmarking, comparative analysis, or large-scale disclosure reviews, Louie's experience offers a useful reference point:
AI unlocks scaled qualitative research analysis, but it is also probabilistic and requires human judgment. The key is staying in the loop: designing workflow assumptions carefully, validating outputs, and being honest about limitations.
Louie's full report is available here. If you want to see more of Louie’s excellent coverage of climate adaptation, check out the Climate Proof website and newsletter, and follow him on LinkedIn.
If you want to try this approach for your own research, brief.green is free to test.
See you next Wednesday!
Natasha and the ClimateAligned team
Explore our products and free tools:
app.climatealigned.co - Access our labelled bond data service with insights on 30,000+ bonds. Check out a free sample!
brief.green - Run AI-powered ESG assessments on-demand. Sign up for free!
AI footprint calculator - Estimate the environmental impact of your AI usage. Try it now!
Project impact web app - Estimate the environmental impact of sustainability projects. Try it now!
Regeneration Atlas - Calculate the climate impact of land regeneration worldwide. Try it now!
Want to learn more about what we do? Visit climatealigned.co
Want to automate a workflow you find annoying and slow? We'd love to hear more! Fill out our Typeform.
