Is your business actually ready to ship AI?
A pragmatic 30-point checklist covering strategy, data, infrastructure, team, and governance — the same one we use in client discovery. Built to surface real readiness gaps, not to flatter you into hiring us.
Five categories. Six questions each.
The projects that ship usually score well on most of these. The ones that stall almost always have a low score in a category they thought was fine.
Strategy & use case clarity
Have you defined what AI is supposed to do — beyond 'use AI'?
Data readiness
Do you have the data, in usable form, with the rights to use it?
Infrastructure & engineering
Can you build, deploy, and keep something running?
Team & ownership
Is there a human who can own this past launch?
Governance, risk & compliance
Do you understand what could go wrong, who's accountable, and your obligations?
Fill it out three ways
Score it yourself, then have two colleagues score it. The disagreements are the signal.
1 · Strategy & use case clarity
- A specific business outcomeHave you identified at least one outcome AI should improve — beyond 'we want to use AI'?
- Measurable successCan you describe success in concrete terms (hours saved, revenue gained, error rate reduced)?
- ROI-ranked use casesHave you ranked candidates by ROI rather than by enthusiasm or visibility?
- An accountable sponsorIs there an executive accountable for the outcome, not just for shipping the project?
- Considered the non-AI pathHave you honestly asked whether the problem could be solved without AI?
- Clear user impactCan you say in one sentence what changes for users when this ships?
2 · Data readiness
- The data exists, machine-readableIs the data the AI will use in a form a system can actually read?
- The data is currentOr is it a 2-year-old snapshot nobody updates?
- You know who owns each sourceAnd have you confirmed they'll give you access?
- Sensitive data is mappedDo you know where PII, PHI, and financial data sits and what your obligations are?
- Labeling, where neededIs your data labeled, or do you need labeling before you can evaluate?
- Enough examples for a golden datasetDo you have 50+ realistic example inputs to evaluate against?
3 · Infrastructure & engineering
- Ownership past the POCDo you have a team or partner who can own the build past the proof-of-concept?
- Architecture decision madeWill AI live in your existing product or as standalone tools — and is the architecture compatible?
- Observability conventionsDo you have monitoring you can extend to AI components, or are you starting from scratch?
- Cost shape modeledHave you modeled per-query vs. fixed-infrastructure cost at 10× current volume?
- A deployment storyCI, staging, rollback — or will every release be a surprise?
- Integration cost assessedDo you know the cost of integrating with your existing systems?
4 · Team & ownership
- Someone who can judge qualityIs there a person who can read AI outputs and judge whether they're good?
- A post-launch ownerDo you have a clear owner for the AI product after launch — not just for the launch?
- Time for adoptionHave you allocated time for end-users to learn and form opinions?
- A feedback loopWill you actually hear about it from production users when something goes wrong?
- Realistic leadership expectationsHas leadership been told what AI will and won't do, including the wrong answers?
- A plan for when AI is wrongCommunication, correction, and escalation paths?
5 · Governance, risk & compliance
- Agreements permit itDo your DPAs, customer contracts, and policies permit the AI use case?
- An autonomy policyIs it documented what AI can decide autonomously vs. what needs a human?
- An audit trail specHave you specified the audit trail you'll need for AI-influenced decisions?
- A wrong-answer answerDo you have a clear response for when AI is wrong — refund, correction, escalation?
- Reputational risk assessedIs the risk of AI being publicly wrong acceptable to leadership?
- Regulatory obligations reviewedGDPR, HIPAA, SOC2, SEC, sector-specific — confirmed compliant?
What your score means.
Count the boxes you confidently checked. Partial credit doesn't count.
Build-ready
You've done the unsexy work. Ready to scope a real build — most engagements at this level ship on the first attempt.
Workable, with gaps
You can build, but two or three categories need attention first. A Discovery engagement closes the gaps before the build.
Pre-readiness
Earlier than you think. Close the gaps first — usually data, ownership, and use-case clarity — before any AI code.
Ready to talk to a human?
Bring your scored checklist. We'll spend 30 minutes on what you actually need to do next — not on a sales pitch.