Free enterprise field guide
How AI should fit into the enterprise SDLC.
AI now touches every stage of software delivery, but most enterprises still treat it as a coding-tool rollout. This playbook maps where AI creates leverage across the lifecycle, where human judgment must remain, and how Enact governs intent, context, approvals, and handoffs above coding agents.
- A lifecycle map for AI across discovery, planning, build, review, release, and learning
- A clear model for where Enact sits above coding agents
- Risk tiers that determine AI authority and human approval
- Clear roles for product, engineering, security, operators, and reviewers
- Verification layers that go beyond tests and plausible-looking code
- A 30-day enterprise rollout and ready-to-use field checklist
8 pages · PDF · Built for CTOs, VPs Engineering, EMs, platform teams, and product leaders
Enterprise field guide
The Enterprise
AI SDLC Playbook
Discover · Plan · Govern · Build · Verify
Enact · Version 1.0
The operating thesis
Coding is faster. Ambiguity is now the bottleneck.
Coding agents compress implementation time, but they also amplify unclear intent. A missing constraint that once triggered a developer question can now produce an entire plausible implementation before the team notices the mistake in review. Most tickets describe a desired change, not the decisions required to implement it — they omit non-goals, architectural boundaries, unresolved assumptions, acceptance evidence, and the conditions under which an agent must stop. Coding agents fill those gaps. Their output may be coherent and still be wrong.
The new workflow
- Use AI to gather context, decompose the problem, and expose what the request leaves unsaid.
- Resolve consequential ambiguity with product and engineering before generating code.
- Approve a shared, machine-readable implementation plan rather than a private prompt.
- Hand the approved plan to the coding agent with permissions, evidence requirements, and stop conditions.
- Verify the result against approved intent — not only against whether the code compiles.
In the full PDF
What the other pages cover
- Ambiguity compression — a 5-step method (gather, decompose, challenge, specify, package) for turning a request into decisions before it becomes code
- Division of responsibility — what AI may do vs. what humans must decide, across context, requirements, technical approach, risk, and acceptance
- The explicit handoff — a portable instruction format for Codex, Claude Code, OpenCode, or any execution agent
- A worked example — compressing ambiguity on a real checkout/payments ticket, decision by decision
- The field worksheet — a 10-point checklist to run before the next coding agent starts