Answer: Oversight should match consequence, reversibility and uncertainty. Internal drafting may require light review; pricing, customer commitments, payments, destructive actions and safety-related decisions require stronger human authority.

What this means in practice

Internal Assistance

Internal research, drafting and summaries can usually run with light review, because errors are caught before anything leaves the company.

Customer-Facing Communication

Messages to customers, owners or subcontractors should be reviewed or sent under approved rules, with a clear escalation path when the AI is uncertain.

Commercial Commitments

Pricing, bids, change orders and contract terms need explicit human approval. AI can prepare the work but should not commit the company.

Financial Actions

Payments and other financial transactions need human authorization. Payment authorization is not a task to hand to AI.

Authoritative Project Records

Changes to the official project record should go through the authoritative system with approval, so the record stays trustworthy.

Safety/Engineering Decisions

Safety, code and engineering judgments stay with qualified people. AI can organize information for them; it does not decide.

A practical construction test

  1. Define the business outcome.
  2. Measure how the workflow performs today.
  3. Identify what AI is being asked to interpret or decide.
  4. Define what remains under human authority.
  5. Test with real examples and acceptance criteria.
  6. Measure results before expanding scale or autonomy.

Where governance fits

The AI model can change. The company’s operating rules should not change silently with it. Catha’s Governance Cage is designed to preserve the business authority layer above the model and workflow.