Use cases / People and insight
AI Decision Support
Recommendations you can trace from data to decision.
01
Problem
Leaders get conclusions without the evidence behind them. When a decision is questioned later, nobody can reconstruct why.
02
AI capability
Gather evidence, model options and trade-offs, state assumptions and uncertainty, and record the recommendation.
03
Data
The agent works from the context you connect, not from the open internet by default.
- Analytics and forecasts
- Risk records
- Policies and precedent
- Internal and approved external research
04
Agent
Decision Support Agent
Prepares option analyses with assumptions and evidence for a named decision owner.
05
Workflow
- 1. Decision framed
- 2. Gather evidence
- 3. Model options
- 4. State assumptions and uncertainty
- 5. Recommend
- 6. Owner decides
- 7. Outcome recorded
06
Governance
Governance runs inside the agent at runtime: policy changes what it can actually do.
Can
- Read evidence it is granted
- Model options
- State assumptions
- Recommend with sources
Cannot
- Make the decision
- Hide assumptions
- Use unapproved data
- Edit the decision record after the fact
Requires approval
- Any action following a recommendation
- Use of restricted data
Records
Agent identity, Data accessed, Model used, Output, Tools called, Policy applied, Decision, Approval, Action, Outcome.
Suggested starting autonomy: L1 Assist
Start at L1: advice only. The decision owner remains accountable and the record links data, model, recommendation and decision.
07
Outcome
What you measure, so the agent earns more autonomy on evidence:
- Decisions with a complete evidence trail
- Time to prepare a decision pack
- Recommendations later revised and why
- Outcome versus expectation
Questions
Does the agent decide?
No. It recommends. A named owner decides and the decision is recorded.
Can we audit a past decision?
The traceability chain links data, model, agent, decision, action and outcome.