Every "AI readiness assessment" I've seen is 40 pages long and takes 3 months. Here are 5 questions that take 30 minutes and tell you more.

#1: Which process costs you the most in labor hours per month?

If you don't know this number, you're not ready for AI. You're not even ready to have a conversation about AI.

This is the starting point for everything. Not "which process would benefit from AI" — that's a technology-first question that leads to bad decisions. The question is: where is your team spending the most time on repetitive, manual work? Get the actual number. Hours per month. Multiply by loaded cost per hour. That's your ceiling for ROI on any automation project targeting that process.

Most operations leaders can name the process but not the hours. "Invoice processing takes a lot of time" is not an answer. "Invoice processing consumes 340 labor hours per month across 8 people" is an answer.

#2: Can you describe exactly how that process works today, step by step?

This is the blocker. This is where most teams stop.

Ask three different people who do the same process to describe it. You'll get three different answers. That's normal — but it means nobody actually knows how the process works. The process lives in people's heads, with personal variations, undocumented workarounds, and tribal knowledge that's never been written down.

If you can't produce a step-by-step map of the process — including decision points, exceptions, and the parts that happen outside of any software system — you cannot automate it. Full stop. AI doesn't fix undefined processes. It just fails at them faster.

If you're stuck here, the next step isn't buying AI. It's running a process discovery project. That takes 2-4 weeks and produces the map you need.

#3: What percentage of that process is already digital vs. manual or paper?

This determines how hard the implementation will be.

If the process runs entirely within digital systems — data moves from one software tool to another, decisions are tracked in a database, outputs are digital files — then automation is relatively straightforward. You're connecting systems and adding intelligence to the handoffs.

If significant parts of the process involve paper documents, phone calls, physical inspections, or handwritten notes, the complexity goes up. You need document processing, speech-to-text, or other input conversion before any automation can start. That's not a dealbreaker, but it changes the timeline from weeks to months and the budget from five figures to six.

#4: Who owns the outcome of this process?

Not who does the process. Who owns the result.

AI projects without a business owner fail. Every time. The business owner is the person who will define success ("we need to reduce processing time from 48 hours to 8 hours"), make decisions when tradeoffs arise ("we'll accept 90% accuracy on document classification if it means processing same-day"), and champion the project when IT raises concerns about integration.

If nobody owns the outcome — if the process is shared across departments with no single person accountable — then fixing the ownership problem comes before the AI project. Shared ownership means shared responsibility means no responsibility.

#5: What does "good enough" look like?

Perfection kills AI projects.

An AI system that handles 85% of cases correctly and saves 60% of labor time can be built in 8-10 weeks. An AI system that handles 99% of cases correctly takes 18 months of training data collection, edge case engineering, and iterative refinement. The first system pays for itself in 4 months. The second one might never ship.

Define your threshold before you start. What accuracy is acceptable? What error rate can your team tolerate if they're reviewing the AI's output? What percentage of cases is it okay to route back to humans? These aren't technical questions — they're business decisions that need to be made upfront.

The best answer to "what does good enough look like" is specific: "If the system correctly processes 85% of standard invoices without human intervention, and flags the other 15% for manual review, that saves us 200 hours per month and pays for itself in one quarter."

If you answered all five questions clearly, you're ready. Not ready for a 12-month "AI transformation" — ready to build one specific system that solves one specific problem. That's all you need to start.

If you got stuck on #2 — and most teams do — that's fine. It just means the first project isn't an AI project. It's a process mapping project. Do that first. Everything else follows.