How would you describe the specific business problem your AI initiative is intended to solve — and how confident are you that AI is the right solution to that specific problem, as opposed to a process improvement or different technology approach?

• We are exploring what AI could do for us rather than starting from a specific problem to solve.
• We have a precise problem statement with documented evidence that AI is the appropriate approach — not a process change or different technology.
• We haven't formalized the problem statement yet.
• We have a general sense of the problem and feel AI is the right approach.

When AI systems produce outputs, the business value depends on whether the people, workflows, and incentives around those outputs actually change. How far along is your planning for those changes?

• We intend to adapt workflows as we learn what the system produces.
• Our understanding is that AI outputs will drive workflow change automatically.
• We have assigned workflow integration design to a later stage of the initiative.
• We have a formal process integration plan that maps AI outputs to workflow changes required for business impact.

What is your current understanding of the data your AI system will rely on — specifically, whether it is complete, accurate, and representative of the real conditions the system will operate in once deployed?

• We know what data we intend to use but haven't formally assessed its fitness for this use case.
• We have reviewed the data at a high level and believe it is adequate.
• We expect to determine data requirements during development.
• We have completed a formal data assessment covering completeness, accuracy, and representativeness for our specific use case.

Does your organization have historical records from the actual business process your AI system will support — records that include real-world exceptions and unusual cases — or will the system be built primarily using data that reflects how the process works under normal, ideal conditions?

• We have historical data but it primarily captures standard cases; edge case coverage is thin.
• We will use primarily benchmark datasets or data from outside our specific operational context.
• We have limited historical data and will supplement with synthetic or proxy data.
• We have rich historical operational data including exceptions and edge cases that represents actual operational conditions.

When you imagine your AI system operating in production, which description best fits what you intend it to do: generate outputs that a person reviews and acts on, or take actions directly in systems and workflows?

• A mix — some outputs are reviewed, some actions are taken automatically, with specifically designed rules determining which category each output falls into, though the oversight boundaries have not yet been formally defined.
• Actions are taken autonomously in systems and workflows; human review occurs at the aggregate or exception level, not for individual actions.
• We haven't yet defined how people will review and act on what the system produces.
• Each significant output will be reviewed by a person before taking effect.

For the people and teams whose daily work will change when your AI system is in production — what has been done to engage them in understanding what the system will ask of them and preparing them for how their role changes?

• We plan to introduce the AI system and communicate the changes when it goes live.
• We haven't started engaging the affected people yet.
• We have had initial conversations with affected teams but haven't formalized preparation or training.
• We have a structured engagement plan that includes the affected people in design decisions, prepares them for role changes, and establishes how they will be supported through the transition.

How would you describe executive sponsor engagement with your AI initiative at this point?

• The executive sponsor is supportive but primarily in a budget approval capacity — day-to-day initiative ownership is below the sponsor level.
• Executive sponsorship is aspirational — we have approval to proceed but haven't secured active engagement from a specific executive.
• We are proceeding without a formally designated executive sponsor.
• The executive sponsor is actively engaged — participating in design decisions, has cleared calendar for the initiative, and has expressed willingness to make the organizational changes required for the system to succeed.

When your AI system is in production and produces an unexpected or incorrect output — one that could cause harm if acted upon — what is your current plan for catching that output and preventing it from taking effect?

• We plan to monitor for problems after the system is live and respond when we detect them.
• We intend to use the pilot phase to define how bad outputs will be caught and stopped.
• We expect the system's accuracy will be high enough that reviewing individual outputs isn't necessary.
• We have a specific plan that includes review checkpoints, criteria for flagging outputs, and a defined path for what happens when something is flagged.

If your AI initiative involves procuring a system or model from a vendor — including foundation model APIs — what obligations have you established with that vendor around transparency, performance, and your ability to govern the system in production?

• We are using a vendor system or model under standard commercial terms.
• We haven't yet engaged vendor governance requirements in our procurement process.
• We are building entirely internally without any vendor-supplied model, API, or data service — this question doesn't apply.
• We have contractual obligations covering transparency about how the system was built and how it changes over time, performance standards, and our right to audit and modify system behavior.

Which of the following best describes your organization's current thinking about AI governance investment?

• We approach AI governance from a strong compliance foundation — meeting regulatory requirements and managing liability is how we ensure the organization is protected and trusted.
• We've found that governance requirements frequently delay deployment at the cost of competitive timing — our priority is moving quickly while meeting minimum compliance requirements.
• We believe governance investment should be proportionate to the risk and value of each AI system — some systems warrant significant investment, others don't.
• We see governance and performance as the same thing — systems built with governance embedded from the start simply work better and return more value than systems where governance is added later.