Implementation decision asset
A first agent is ready when the operating contract is clear.
Use this checklist before approving build work. It covers the decisions that determine whether an agent can act safely, be tested honestly, and produce a measurable result inside an existing team.
Pre-deployment checklist
Four gates to clear before production access.
Workflow and ownership
The agent needs a bounded role rather than a broad instruction to help the team.
- Name the trigger, inputs, expected outcome, and excluded actions
- Document normal and exception paths
- Assign policy, data, system, and human-escalation owners
- Define who can pause, change, and accept the workflow
Data and system access
Access should be minimal, attributable, and usable in the intended environment.
- Identify approved sources and destination systems
- Validate API, sandbox, credential, rate-limit, and field access
- Review quality, duplication, missing values, and source conflicts
- Set retention, deletion, credential rotation, and audit-log expectations
Guardrails and testing
Failure paths must be designed before launch, not discovered through customers.
- Define approval-gated and prohibited actions
- Set uncertainty, confidence, retry, and human-takeover rules
- Test normal, missing, duplicate, delayed, conflicting, and unavailable inputs
- Use realistic acceptance cases and preserve evidence
Measurement and operations
Launch needs a baseline, owner, review rhythm, and cost view.
- Record current volume, handling time, delay, errors, and backlog
- Choose outcome, quality, guardrail, and cost metrics
- Forecast model, channel, infrastructure, licence, and support cost
- Schedule launch monitoring and a 30-day operating review
Buying model comparison
Readiness is evidence—not enthusiasm.
A short discovery or blueprint can close important gaps before deployment.
Workload
Ready to scope deploymentOne recurring, bounded, measurable operating queue.
Needs discovery firstBroad request to automate a role or department.
Systems
Ready to scope deploymentAccess path and source-of-truth decisions are known.
Needs discovery firstAPIs, credentials, data quality, or ownership are uncertain.
Risk
Ready to scope deploymentApprovals, fallbacks, prohibited actions, and stop owner are defined.
Needs discovery firstExceptions are handled informally or no human queue exists.
Evidence
Ready to scope deploymentBaseline and acceptance thresholds can be measured.
Needs discovery firstSuccess is described only as saving time or using AI.
Readiness result
Proceed only when the critical gates are owned.
Not every detail must be perfect, but each unknown needs an owner, decision date, and effect on scope.
Proceed to scope
- One workflow and business owner are named
- Required systems and fields are accessible
- Human exceptions and stop conditions are defined
- A baseline and acceptance test can be produced
Pause and discover
- No agreed source of truth exists
- The workflow depends on unknown integrations
- Irreversible actions have no approval route
- Success has no measurable baseline
Buyer objections
Questions to resolve before signing.
Do we need every checklist item for a read-only internal agent?
The depth changes with risk, but ownership, source data, access, failure behaviour, testing, and measurement still matter. A read-only role may require lighter approvals than customer messaging or record changes.
What if our API access is uncertain?
Validate it before committing to a delivery schedule. A Workflow Blueprint can map alternatives, but unavailable access can change feasibility, cost, and timing.
Can requirements evolve during delivery?
Yes through controlled change decisions. The accepted workload and success criteria should remain stable enough to test; material new systems or roles need re-scoping.
Continue the evaluation
Continue your deployment evaluation.
No-pressure scope check
Check the workload before implementation starts.
The ROI audit identifies whether the selected workload justifies a deployment conversation, a narrower discovery, or no project.
Start the Agent ROI Audit