APEX-Agents category
AI Agents for ESG and Climate Risk Analysis
This page showcases APEX-Agents tasks that test whether AI agents can analyze ESG and climate risk, including CDP data, climate risk scores, and portfolio reallocation decisions.
Primary tasks
3 tasks with this category as their main focus.
-
How much cash in EUR will adopting Rakling's strong sell recommendations generate for HP? Assume they sell the position as of October 30, 2025 at no cost. Report your final answer here in dollars and cents.
Expected output: message_in_console -
Based on Planet Defense's climate risk data and their scoring methodology, give me the top 3 sub-regions by overall risk score and their overall risk scores. Once you've done that, calculate for each continent the average overall risk score and the standard deviation of all the overall risk scores. Report numeric final answers to 2 decimal places. Write your answer here.
Expected output: message_in_console -
Please check how KO and MDLZ differ in expected upside once we apply the ESG and GLP-1 filters and account for each investor’s maximum allowable ESG risk level. Use the survey data and the ESG risk thresholds to determine which respondents are eligible to hold each company. Then calculate the confidence weighted average and standard deviation of expected annual return for KO and MDLZ. Show each company’s weighted average and standard deviation of expected annual returns. Round only the final results, going to two decimal places.
Expected output: message_in_console
Related tasks
7 tasks that also exercise this type of work as part of a broader assignment.
-
Investigate whether EuroGrid should consider increasing staffing. Determine if the number of working people per impacted asset is correlated with the expected economic impact of unforeseen downtime in each Country-Region combination. Assume that downtime also includes emergency repairs. Let's conduct 2 regression analyses using data in each country-region pair: - [Workers Per Asset] vs [Economic Cost Per Worker Per Weather Event] for weather related outages - [Workers Per Asset] vs [AVG Emergency Repair Cost]. Provide the R² value for each relationship to the nearest 2 decimal places. More investigation is warranted so long as both models have R² value > 0.5. Based on the models, recommend whether to proceed with this investigation or not. Keep this in mind: - For each analysis, use unique asset counts that correspond to the underlying dataset used when calculating workers per asset. - For both assessments we can assume that all workers in the workforce are supporting responses to unforeseen downtime and that workforce size has not changed in the past 5 years. - For emergency repair costs, use the simple average of the annual repair cost over the full 5 year history (2020 - 2024) for each country-region pair. - For each individual regression analysis only use the data present in both sets of data needed for that regression (e.g., if Austria Alpine has workforce data and weather data but no emergency data then it will be used in the 1st regression but removed from the 2nd regression analysis). -Use the EuroGrid's maintenance CapEx/OpEx 5-yr summary file to get the emergency repair cost figures for each country-region pair. Use the Grid workforce and maintenance productivity file to get workforce size. Use the extreme weather and climate stress dataset to get the number of impacted assets and total weather events per year. Write out the answer for me here in a brief message.
Expected output: message_in_console -
Update numbers with the new projections (attached). I want the full breakdown for: DC converters and onboard chargers Driveline and axle modules Engine control units Engine core hardware Exhaust and emissions Fuel and injection systems On vehicle charging hardware Power electronics and inverters Sensors and wiring Structural EV content Thermal management modules Transmission and e drive Ignore sensors and structural EV content. Round final numbers to two decimals, and reply just straight back in here.
Expected output: message_in_console -
Given our current set of CDP questionnaire responses for Horizon's portfolio companies, identify the three largest risks by potential financial impact across all entries. In cases where a minimum and maximum impact are listed for an individual risk, use the midpoint. For risks where there is no impact figure listed, assume it is zero. If two risks are tied in terms of dollar impact, prioritize the more recent risk as of 2024. Reply back to me, outlining the 3 Risks (defined as the combination of the company, year, and type of risk), the Financial Impact, and the Mitigation Ratio (defined as the cost to correct the risk divided by the potential financial impact). Give figures in the currency indicated in section C0.4 of their associate CDP response file. Round Mitigation Ratio to the nearest 0.01, and give long form currency values.
Expected output: message_in_console -
I feel good about our current assessment of the valuation, but I’d like to do some forward-looking assessments. Can you use the historical Sector Median PE volatility data to determine which of the currently undervalued stocks are at the highest risk of becoming overvalued. Give me the company name, ticker symbol, and the probability percentage. Just to reiterate, a premium of 25% or more over the Sector Median PE is considered overvalued. Anything else is undervalued. Use 28.5 as the current Sector Median PE. I think it’s fair to assume the same PE volatility distribution will continue. Round final percentage to two decimal places. Reply back to me with your answer.
Expected output: message_in_console -
Can you calculate the minimum gross returns Crown and Vision would have to achieve in order to deliver net returns equal to the average fund manager in the 80th percentile or better in terms of Sharpe ratio? Report numeric final answers to two decimal points. Write your answers out here.
Expected output: message_in_console -
Calculate the CAGRs for the ABInBev's 2025 sustainability goals, starting with the 2021 results. Some goals imply declining metrics, like water use, while others are looking to increase, such as the use of renewable electricity. Accordingly, provide an average for each of the top two most positive CAGRs and the two most negative CAGRs. These should be taken as the target CAGR for all Sustainability Goals that need to increase and decrease, respectively. Next, use these two CAGRs to determine the year that each goal would be achieved. Only evaluate the first 14 sustainability goals listed. Also, consider only goals with a defined 2025 target in the report. Round all final results to two decimal places, and display years as whole numbers. Please give me your answer here as a reply.
Expected output: message_in_console -
The client is looking to execute our rebalancing recommendations. Identify the stock with the highest Absolute Beta Reliability during the worst 10 trading days in Q3 2025, defined as the worst 10-day cumulative return of the F&B Sector Index. For each stock, compute: - Beta Predicted Return = Beta Coefficient × 10-Day Return of the F&B Sector Index - Absolute Deviation = |Stock 10-Day Return − Beta Predicted Return| - Absolute Beta Reliability = 1 − (Absolute Deviation ÷ Stock 10-Day Return) Reply back to me in a message with the company name, and the highest Absolute Beta Reliability and its Absolute Beta Reliability value. Round numbers to four decimal places.
Expected output: message_in_console