Decision reasoning

Compass
Make the options and reasoning visible.

Compare options and produce explicit recommendations from operational context, deterministic logic, and approved intelligence signals.

An illustrative decision record

Compare responses before committing.

A delivery exception becomes an explicit comparison of options, evidence, and tradeoffs for the person or policy responsible.

Situation
A carrier delay puts a customer commitment at risk.
Options
Keep the route, use an alternate carrier, or revise the commitment.
Tradeoffs
Compare delivery timing, available capacity, cost, and risk.
Next step
Send the recommendation through Policy and Workflow.

Illustrative workflow view. Controls and integrations depend on deployment.

Evaluate one decision

Use operational context to identify candidate responses and compare the consequences of each. Make the baseline and proposed options inspectable.

  • Candidate options
  • Decision-local outcome projections
  • Cost, risk, and service tradeoffs

Use approved intelligence

Bring deterministic logic, approved model signals, and active pattern findings into the comparison when the operating decision calls for them.

  • Deterministic decision logic
  • Scoped model signals
  • Relevant pattern findings

Explain the recommendation

Record the evidence, options, recommendation, and uncertainty in a DecisionCard. Policy gates the proposal and Workflow records the authority to proceed.

  • Evidence and rationale
  • Confidence and uncertainty
  • Policy and authority references

In practice

Make consequential options easier to review.

Compass is for operating decisions that need context, tradeoffs, or an explanation. It structures the comparison so the person responsible can understand the proposed response.

Delivery recovery

Compare keeping a route, using an alternate carrier, or revising a commitment. Put available capacity, timing, cost, and customer impact alongside each response before it moves into the control path.

Maintenance priorities

Evaluate responses to a production exception using the accepted equipment and work context. Show the baseline, constraints, and implications of the proposed response for operational review.

Resource coordination

Compare reassignment and escalation options when capacity changes. Keep the evidence and uncertainty visible so an operator can judge whether the recommendation is ready for approval.

How it works

Give the recommendation a reviewable record.

The DecisionCard connects the evidence, options, recommendation, and control path. It provides a clear handoff from reasoning to authority.

  1. 01

    Frame the decision

    Identify the operating question and relevant context. Define the baseline and candidate responses so the comparison answers a specific decision rather than an open-ended prompt.

  2. 02

    Compare the responses

    Use deterministic logic, approved model signals, and relevant findings to assess the options. Preserve the assumptions, tradeoffs, and uncertainty behind the ranking.

  3. 03

    Hand off for control

    Record the recommendation with its evidence and rationale. Policy determines whether it is allowed, blocked, or approval-required; Workflow preserves the authority path.

Common questions

Understand the fit.

Does every workflow need an AI model?

No. Decisions can use deterministic logic and rules. Approved model signals are used when appropriate to the operating problem, available evidence, and control requirements.

Can Compass execute its own recommendation?

No. Reasoning and execution have separate responsibilities. A recommendation must pass through the defined policy and authority path before Praxis can execute it.

How can an operator review uncertainty?

The DecisionCard exposes the evidence, candidate options, rationale, and relevant uncertainty. The workflow defines when missing evidence or unresolved tradeoffs require human review or escalation.

Start where one workflow needs clearer control.

Adoption starts with one workflow, the systems around it, the approvals it needs, and the outcome your team wants to improve.

Design a decision loop