Answer questions to get your recommendation
Start the assessment below to see which approach fits your development task.
What's the complexity of your task?
Decision Framework
Complete the assessment above to see how your answers map to our proven decision framework.
| Factor | AI-First | AI-Assisted | Agent-Delegated | Human-First |
|---|---|---|---|---|
| Code Complexity | Simple patterns, boilerplate | Standard business logic | Repetitive tasks with clear acceptance | Novel algorithms, complex architecture |
| Security Risk | Low (UI, docs, tests) | Medium (business logic) | Guarded scopes, low blast radius | High (auth, payments, crypto) |
| Team Experience | Senior devs who can review | Mixed experience teams | Confident in reviewing agent output | Junior devs learning concepts |
| Context Available | Similar code exists in codebase | Some existing patterns | AGENTS.md + skills + tests ready | No precedent or patterns |
| Time Constraints | Tight deadlines, standard work | Balanced delivery timeline | Async via Copilot app session, cloud agent, or CLI /fleet | Learning/exploration priority |
| Agent Autonomy Risk | N/A | Low autonomy, close supervision | Custom agent scopes, sandbox policy, blast-radius limits | Requires real-time human judgment |
| MCP Integration Available | Optional | Useful but not required | MCP servers on the enterprise allowlist | Not applicable |
| Tooling | Copilot inline completions, next edit suggestions | Copilot chat, agent mode, Spaces | Copilot app sessions, cloud agent, Copilot CLI /fleet, custom agents | IDE + manual workflows |
| Verification | Tests + Copilot code review | Rubber Duck second opinion + human review | Agent self-review, security review, Agent Merge conditions | Pair review + design review |