Scenario Planning Is Decision Rehearsal
Scenario planning is decision rehearsal
Scenario planning is useful when a team has to decide before the portfolio has
revealed the outcome.
The useful output is a visible assumption set before money, risk, or operational
effort moves.
A forecast says, "this is what will happen." A scenario says, "if these
assumptions hold, this is what the decision may do." That difference matters.
One creates false certainty. The other gives a team something to argue with.
The product rule
An AI analyst can help with scenario planning when it stays close to the
evidence:
- start from the live portfolio state;
- name the historical behavior or user-supplied assumption being applied;
- show the base population and exposure;
- show sensitivity to the assumption;
- label the result as a planning estimate; and
- keep the data, method, and caveats traceable.
That is the guardrail. A scenario answer without visible assumptions is just a
confident guess with numbers attached.
Why this belongs near BI
Dashboards show the current state. Analysts help a team decide what to do about
it.
Scenario planning is the bridge between those two jobs. It turns "PAR30 is up"
into "what would we have to collect, pause, tighten, or watch for this decision
to work?"
This is still Interrogable Business Intelligence. The follow-up question is
now forward-looking, so the system has to be stricter about what is fact, what is
assumption, and what is judgment.
Notto code truth
Notto AI Agent has current support for bounded what-if questions. The
forecasting layer would be separate product and governance work.
Verified on 2026-08-13:
packages/ai-agent-microfrontend/src/lib/starter-questions.tsincludes
what-if prompts such as DPD roll-forward and collection needed to stay inside
investor thresholds.packages/ai-agent-microfrontend/src/components/chat/workflows/workflow-templates.ts
includescounterfactual-cure-simulator, which queries current delinquent
exposure and asks for a PAR30 cure plan.- The template explicitly frames the output as a planning estimate rather than a
model forecast. - Missing today: approved macro scenario engine, calibrated loss model,
backtesting harness, and capital stress-testing workflow.
That is a good boundary. The current product can rehearse a decision. Forecasting
the loan book would require more evidence and governance than the repo has today.
Research check
This idea is still valid beyond Notto.
The OCC treats scenario or sensitivity analysis and stress testing as normal
credit-risk management practice, with the method scaled to the bank's portfolio
strategy, size, loan types, composition, operations, and management:
https://www.occ.gov/news-issuances/bulletins/2012/bulletin-2012-33.html
The Federal Reserve's interagency stress-testing guidance frames scenario
analysis as applying historical or hypothetical scenarios to understand impact.
It also stresses input data, assumptions, uncertainty, limitations, and
documentation:
https://www.federalreserve.gov/frrs/guidance/interagency-supervisory-guidance-on-stress-testing-for-banking-organizations-with-total-consolidated-assets-of-more-than-10.htm
The lesson I want to keep is smaller than bank stress testing: a good product
can offer decision rehearsal before it earns the right to call itself a
forecasting system.