What problem is worth solving?
Teaching scenario: sales, product and operations disagree about a fictional market entry. Expose the conditions behind each view and the order of validation rather than smoothing away disagreement.
Help decision-makers identify missing information and evidence that could change the choice.
- WORKING ROLES
- Decision owner, domain research, product/operations and FDE
How does an FDE turn symptoms into a solvable problem?
The following discovery and implementation exercise uses a synthetic scenario.
Define the decision
Specify whether to enter, which segment and what validation budget. Horizon and reversibility determine research depth.
Expose hidden assumptions
List assumptions and failure conditions per perspective; map conflict to price, cost and demand rather than voting.
How does the solution enter the workflow?
- 01
Decompose
Break down demand, competition, channel, delivery and economics.
- 02
Retrieve
Keep dated evidence and counterarguments; distinguish observations from forecasts.
- 03
Compare
Define conditions for entry, delay and a limited test.
- 04
Validate
Select a cost-effective information test and a decision-changing threshold.
Which decisions make the solution reliable?
Perspectives are not expert testimony
Role prompts structure analysis, not expert testimony; real expert claims need real sources.
Keep the counterargument
Name the weakest assumption; limited trials or delay can be valid recommendations.
Entry and delay can both be rational
Fictional project: price 100, delivery cost 60–110, with interest interviews but no purchase evidence.
- Entry condition
- Costs stay below price with evidence of payment.
- Reason to delay
- The cost ceiling exceeds revenue; interest is not an order.
- Next validation
- Run a bounded delivery trial and record effort and willingness to pay.
What should you do and deliver?
Define the decision and exclusions
- OUTPUT
- Decision brief
- PASS CONDITION
- Questions can change a decision
Link facts and perspectives
- OUTPUT
- Evidence matrix
- PASS CONDITION
- Conflicting evidence remains visible
Define tests and exit conditions
- OUTPUT
- Three options + test plan
- PASS CONDITION
- Later evidence can falsify the recommendation
How do you prove completion and reuse the work?
Acceptance checks
Review source quality, counterargument, assumptions and executable validation, not report length or role count.
Reusable assets
Reuse decision trees, evidence matrices and conditional recommendations for entry, selection and investment.
Further reading and original sources
- Building effective agents ↗Anthropic · 2024-12-19
- Demystifying evals for AI agents ↗Anthropic · 2026-01-09