AI Readiness Assessment: 25-Question Self-Scored Scorecard

Each section has 5 questions worth 0–4 points (0 = not at all, 4 = fully in place). Maximum score per section is 20; maximum total is 100. Score yourself honestly — inflating numbers here means buying software that fails three months later. After scoring, jump to the 'Score Bands' section to read what your range means and which gaps to close first.

1. We can name the specific business problem AI will solve in one sentence — without using the words 'AI', 'automation', or 'efficiency'. 2. We have measured the current cost or pain of this problem in dollars, hours, or error rates — not just 'it's bad'. 3. We know what 'good enough' looks like — a specific target number we'd be happy hitting. 4. We have ruled out non-AI fixes (better process, better hire, better software) and AI is genuinely the right tool. 5. The problem is recurring and high-volume enough that automating it is worth the setup cost.

6. The data the AI will need is stored somewhere we can access — not trapped in PDFs, paper, or one person's head. 7. The data is reasonably clean and consistent — same fields filled in the same way across records. 8. We have at least 6–12 months of historical data for the problem we're solving (if pattern recognition is involved). 9. We know who owns the data, who can grant access, and what compliance rules apply (HIPAA, GDPR, CCPA, industry-specific). 10. We're comfortable with where the data will go — on-prem only, private cloud, or shared with an API provider — and have a written stance.

11. There is one named person inside the company who will own the AI project day-to-day — not a committee, one person. 12. That person has 5+ hours per week reliably available to spend on the project. 13. The team that will use the AI day-to-day has been told it's coming and isn't actively opposed. 14. Someone on the team (or a trusted contractor) can read and interpret outputs critically — we won't be flying blind on whether the AI is right or wrong. 15. We have or can buy access to someone who understands AI implementation — internal hire, consultant, or vendor partner — and we've budgeted for it.

16. An executive sponsor at the company has personally said 'I want this' — not 'sure, try it'. 17. The budget covers initial implementation AND ongoing costs for 12 months (API usage, maintenance, retraining/tuning, additional licenses). 18. We've reserved 15–20% of the budget as contingency for the surprises that always show up in phase 2. 19. Leadership has agreed on a clear go/no-go date — a pilot won't quietly drift for 18 months without a decision. 20. There is patience for a pilot phase where results are mediocre while the system is tuned — month one is rarely the win month.

21. We have a 'what if it's wrong' plan — a human review step, an escalation path, or a way to roll back. 22. We know what we'll tell customers, employees, or regulators about how AI is used (disclosure language, opt-outs, audit trail). 23. We have a monitoring plan — someone will look at performance numbers weekly during the pilot, not just at the end. 24. We have an exit plan if the vendor disappears, raises prices, or shuts down the product — our data and process aren't trapped. 25. We've thought about what we'll do when the underlying model gets meaningfully better in 6–12 months (frontier model providers ship upgrades that often) — pinning to today's model is a real risk.

**0–40 (Not Ready):** Don't sign a contract yet. The biggest gaps are usually in Sections 1 and 2 — problem definition and data. Spend 30–60 days fixing those before you talk to vendors. The cost of a failed AI implementation is 5–10x the cost of waiting two months. **41–65 (Almost Ready):** You can pilot, but pick one narrow use case. Don't start with the company-wide vision; start with the highest-scoring section's lowest-friction problem. Use the pilot to fix the gaps the score revealed. **66–85 (Ready to Pilot):** You're in good shape. Move forward, but use a 90-day pilot with explicit success criteria. Re-score after the pilot — readiness changes as you learn what production actually looks like. **86–100 (Ready to Scale):** Rare. Either you've already done this before, or you're inflating scores. Have a second person score independently and reconcile differences. If the score holds up, you're cleared to skip the pilot phase on small/medium use cases and go straight to a phased rollout.

If your lowest-scoring section is **Problem Definition**, slow down. Re-run the problem in plain English with five people who'd be affected, and write down their answers. AI cannot solve a problem the company hasn't agreed on. If your lowest-scoring section is **Data Readiness**, the fix is unglamorous: hire someone for 2–4 weeks to inventory and clean the data. This is the single most common cause of stalled AI projects, and the fix is operational, not technical. If your lowest-scoring section is **Team & Skills**, name the owner before anything else. AI projects without a single human accountable will not ship. If nobody internally can own it, hire a fractional AI lead or partner with a consultant who'll act as one. If your lowest-scoring section is **Leadership & Budget**, get the sponsor conversation handled before spending another dollar. Pilots without active sponsorship are politically vulnerable when results dip in week 6, which they always do. If your lowest-scoring section is **Operations & Risk**, build the rollback and monitoring plan first — before the model goes live. Companies that skip this discover the gaps the first time the AI is wrong in front of a customer.