Agent Marketplace · Strategy to Plan
Scenario Planner Agent
Builds best, base and worst case scenarios from your financial and market inputs.
How to use this agent
Stage 1 of 5
See what it needsRead the input data the agent asks for and where each figure usually comes from.
Stage 2 of 5
Run the demoPress the demo button to see a full run on a fictional company. Nothing is saved.
Stage 3 of 5
Get the templateThe tool opens once it is approved; the template downloads from there.
Stage 4 of 5
Add your dataFill the template and upload it, or type straight into the tool. Your work saves as you go.
Stage 5 of 5
Read and share the resultRead the headline results, then print or save the PDF report.
SIPOC: how the agent works
Who supplies what, what goes in, what the agent does, what comes out, and who uses it.
| Suppliers | Inputs | Process | Outputs | Customers |
|---|---|---|---|---|
|
|
|
|
|
Input data it needs
Everything the agent asks for, in the order it asks. Columns marked "worked out" are calculated for you; you only fill the rest.
| Data | Unit | What to enter |
|---|---|---|
| What are you deciding? | Describe the choice in one or two sentences, including the date by which it must be made and who makes it. | |
| What is the payoff measured in? | Name the single measure every payoff uses, with its unit and period, so that all options are compared like with like. | |
| What limits the choice? | Name the budget, capacity, regulation or time limits that make some options impossible, so the options below are all feasible. | |
| Likelihood of the best case | % | Your honest probability that the favourable scenario happens. The three likelihoods should total 100; if they do not, the maths rescales them and a warning shows. |
| Likelihood of the base case | % | The probability of the most likely, middle scenario. Use the same evidence for all three likelihoods, such as order book, market data and expert views. |
| Likelihood of the worst case | % | The probability of the unfavourable scenario. Do not set this to zero: if it cannot happen, it is not the worst case you should plan for. |
| How much to penalise risk | x spread | Each option is scored as expected value minus this number times its spread. Use 0 if the organisation can absorb a bad outcome, 0.5 for a balanced view, 1 or more if a bad outcome would be severe. |
| Optimism index for the blended rule | index 0 to 1 | Used only by the Hurwicz rule: 0 trusts only the worst case, 1 trusts only the best case, 0.5 weights them equally. Choose before you see the results, so the answer cannot steer the choice. |
| Optionstable | One row per option, including the option of doing nothing. Use the same unit for every payoff and include all costs, so a loss is a negative number. Columns: Option; Payoff if best case (in your unit); Payoff if base case (in your unit); Payoff if worst case (in your unit); Expected value (worked out); Spread (risk) (worked out); Risk-adjusted score (worked out); Worst case (maximin) (worked out); Best case (maximax) (worked out); Blended (Hurwicz) (worked out); Regret score (worked out). | |
| Early-warning indicatorstable | One row per indicator you can observe before the outcome is known. A good signpost has a number and a threshold, not an opinion. Columns: Indicator; Threshold; Signals which case; Response if it trips. | |
| What did you decide, and why? | Write the decision with the main reason and which rules supported it, so a person who was not in the room can follow it later. | |
| Actions for the next 90 daystable | List the first actions in order. Give each an owner by role, a measure that shows it worked, and when it starts. Columns: Action; Owner (role); Measure of success; When. |
Try the demo
See a complete run on a fictional company before you enter anything of your own. No account, no AI, nothing saved.
Demo · sample, not real
A packaging manufacturer must decide on capacity for the next three years: expand its own plant, outsource overflow, or hold. Payoffs are cumulative operating profit uplift over 3 years in ₹ lakh. This is a fictional example for learning.
Provide and update your data
- Open the pilot and choose Download blank Excel template, or Download sample to see a filled example.
- Fill the template with your own data and use Upload to load it; or type straight into the page. Both give the same result.
- Your work saves automatically while you type, so you can come back and change any figure; results update at once.
- To refresh: Update the likelihoods and payoffs whenever a signpost moves or at each quarterly review.
The pilot is being reviewed and opens to clients once it is approved.
What you get back
Headline results
- Highest expected value
- Best after allowing for risk
- Safest (best worst case)
- Boldest (best best case)
- Best on your optimism blend
- Least regret
Working figures shown beside them
- Likelihoods add up to
- Likelihood check
- Expected value with perfect information
- Value of perfect information
In what form
- Results at a glance on the page, updating as you type. Available
- PDF report in the standard Eccellere layout, from Print / PDF on the tool page. Available
- Dashboard over time that keeps each run and charts the change between runs. Planned
The agent never lets AI invent a number: every result comes from the formulas below, and any AI writing is labelled and editable.
Method and limits
Based on
Abraham Wald (maximin), Leonard J. Savage (minimax regret), Leonid Hurwicz (optimism index), Howard Raiffa and Robert Schlaifer (expected value of perfect information); scenario thinking after Peter Schwartz, The Art of the Long View (1950 to 1991).
A payoff table sets each option against each possible future. Expected value averages the payoffs by likelihood, the spread measures risk, and four rules for deciding without trusting the probabilities act as stress tests.
Use it when
Use it when you must commit capital, capacity or effort before the future is known, and you have two or more options and a view of how likely each scenario is.
Do not use it when
Do not use it when only one option exists, when the outcomes cannot be put in the same unit, or when a legal or safety duty decides the matter whatever the payoff.