Business AI Implementation Readiness Checklist

By KODE team · 25 September 2026

Start with one task

Choose one process with identifiable users and an owner. Complete the checklist with people who understand the daily work, source data, and required approvals. An Unknown status is useful: assign an owner and a verification step to each uncertainty before turning it into a delivery commitment.

Readiness discussion checklist

Tick items as reading notes. Your selections are not sent to or saved by KODE. Use your browser’s Print feature to save a PDF copy.

• The goal and users are clear
  Write down the supported task, intended users, decision being supported, and what is outside the pilot.
• There is a baseline
  Record current quality and task time, including review and corrections. Agree a comparable way to measure change.
• Data owners and usage permissions are known
  List sources, owners, classifications, and permitted uses. Use sanitized examples in introductory discussions.
• Documents and figures have an authoritative version
  Identify approved sources, active versions, metric formulas, and the process for resolving conflicting information.
• Integration access can be tested
  Confirm that application owners know the integration plan. Define a test environment and scoped access without sharing personal credentials.
• Action boundaries are defined
  Separate reading, drafting, and modifying data. Specify human approvals, prohibited actions, and escalation routes.
• Evaluation includes failures
  Prepare normal tasks, ambiguous requests, missing information, denied access, and application failures. Define expected behavior.
• Acceptance criteria are agreed
  Set context-appropriate quality requirements, assessors, evaluation evidence, and conditions to continue, revise, or stop.
• Costs and usage assumptions can be explained
  Separate delivery, models, infrastructure, and support. State user assumptions and triggers for cost or scope changes.
• Operations and support have owners
  Assign document updates, quality monitoring, incident handling, and user support after launch.
• Users understand how to review outputs
  Prepare task guidance, usage boundaries, example errors, and a way to reach a person when results are unsuitable.
• Exit and handover arrangements are clear
  Agree data export, account and code ownership, access removal, documentation, and a return to the previous workflow if needed.

Use incomplete items as an action list, prioritizing those that prevent evaluation or increase the consequences of failure. The number of checked boxes does not establish readiness. NIST’s AI RMF and Playbook can provide further reading on risk discussions; this checklist is KODE’s practical tool, not an official NIST assessment. To start a conversation with KODE, describe the process, intended outcome, and three uncertainties.