Playbook
The repeatable method
The steps for a recurring task, the context they need, and where a person checks before work moves forward.
The system of record for AI work.
The best way of working with AI becomes a playbook the whole team can run, review, and improve.
What we keep for your team’s AI work.
Playbooksa growing library of the work you repeat
SalesInbox → CRM triage
SupportSegment-stage reply
BizDevRFP first draft
+11
Runshow often each one is performed
147×
Review recordwhat we learned across every run
Versionsversion controls your work
v1v2v3
v3 updated the guideline
What Constraint keeps.
The method, each time it runs, the record of what a human checked, and the version that carries improvements forward.
Run
One time through
A single execution of a playbook’s ordered steps. Each run is its own instance you can point to.
Review record
What a human checked
A short receipt for a run: what was reviewed, the evidence behind it, and who signed off.
Version
The method, improving
The approved shape of the playbook. An accepted change becomes the next version.
Editing AI’s work shouldn’t be harder than redoing it.
But it usually is: a finished output hides where it went wrong, so starting over often wins. Constraint catches the wrong turn early, and every run leaves a record: the steps followed, the context used, who signed off at the gate, and the result produced.
Review record
Weekly report
What was reviewed
The playbook and the step that mattered.
Visible evidence
The context and sources the work used.
Review gate
Who checked it before it moved forward.
Result
The output the run produced.
You can always tell.
Every result shows its work. You can see where each step came from, what it drew on, and why.
Provided
Pulled straight from your own material and cited: the exact steps your SOPs and docs called for.
Approved context
What’s true about your business but never written down step-by-step: your knowledge files, team structure, what you sell.
Guidance
Where AI filled a gap with its own research to finish the job, clearly marked as its own, so it never hides in with the rest.
AI hands you an answer and hopes you trust it. Constraint shows you what went into it, so trust isn’t a leap of faith.
The same quality, whoever runs it.
The same method, every time.
Repeatable quality
Everyone starts from the same vetted method, so good work doesn’t depend on who is at the keyboard.
It improves
Approved changes sharpen the method with use, instead of resetting from scratch each time.
It expands
Take on work the team hasn’t run before: start from the Bank’s vetted method instead of a blank page.
See the process, not the chats.
Constraint gives the people accountable for AI-assisted work a clear view of the method and its review trail.
Baseline
Which recurring workflows are being adopted, and where AI hasn’t landed yet.
Correct use
How the work is actually run: what was reviewed, and who signed off.
Version up
Each approved improvement becomes the version the whole team runs next.
Companion, not another tool
What Constraint manages, and what stays outside.
Constraint sits beside the AI your team already uses, not another tool to buy.
Constraint manages
The shared playbook, review gates, review records, and version history.
Outside Constraint
The model, its tools, and the files it can reach. That stays in your AI environment.
- ChatGPT
- Codex
- Claude.ai
- Microsoft Copilot
- Google Antigravity
- OpenClaw
Connected through MCP, the standard plug between AI tools and other systems. No migration, no new login.
Constraint also ships a Bank of expert playbooks distilled from 1,100+ operators, via Firneo. Vetted starting points you can adapt to your team.
Browse the BankStart with one workflow.
Bring one recurring AI workflow. We’ll show how it becomes a shared, reviewable playbook.