Checklist: is your company ready for AI implementation

A conversation about AI implementation usually starts with a question about the model or the vendor. The more useful question is different: is this specific process, with this specific data, at this specific company, actually ready for automation. The checklist below lets you check that yourself, without writing a single line of integration code.

How to use this checklist

This isn’t a test you need to pass at a hundred percent before you can even start talking about AI. It covers the five areas we check during an AI-readiness process audit: process, data, business owner, maintenance budget, and success criteria. For each item, mark an honest yes or no. The more yeses, the sooner a given process is ready for a pilot. A no isn’t a failure — it just shows what’s worth sorting out before you decide to start.

Process

• The process is repeatable enough that its typical path can be described step by step.
• You know the exceptions to the standard path — not just the documented version, but what actually happens.
• Someone at the company can describe the process as it really runs, not just how it’s supposed to run.
• The process has a clear start and end, rather than blurring across several departments with no single owner.
• You know how long the process takes today, so there’s something to compare the result against after implementation.

Data

• The input data exists in electronic form, or can quickly get there.
• You have samples of the documents or data to check whether AI can actually handle them.
• The data format is reasonably consistent — not every document looking different and coming from a different source.
• It’s clear where that data lives today and who has access to it.
• The data doesn’t need a separate, large cleanup project before AI can read any of it.

Business owner

• There’s a person on the company side who owns the outcome of the project, not just someone cheering from the sidelines.
• That person has time to take part in conversations about the process, not just sign off on an invoice at the end.
• The team that currently runs the process manually knows about the automation plans and will be involved in testing.
• The business owner can decide what happens to cases that get escalated.
• The company has one decision-maker, not several decision centers that disagree with each other.

Maintenance budget

• The budget covers not just the pilot and the rollout, but maintenance after launch.
• Someone understands that maintenance is a recurring cost, similar to hosting or licensing, not a one-off line item.
• The company assumes document formats and systems will change over time, and the implementation will need adjustments.
• There’s a plan for who monitors the system after launch — an internal team or a vendor.
• The budget doesn’t assume AI will work flawlessly from day one with no fixes needed.

Success criteria

• The company can state upfront what “success” for this implementation actually means — concretely, not vaguely.
• It’s possible to measure how much time or effort the process takes today, to compare against the result after implementation.
• It’s clear what the cost of error is for this process, and who bears it if AI gets something wrong.
• It’s clear at which point AI proposes a result and at which point a person or a deterministic rule decides — the same principle described in “AI proposes, code decides”.
• The company accepts that a pilot might show the process isn’t actually ready for automation yet — and that’s a valuable outcome too.

Why these five areas specifically

Each of these five points corresponds to a different way an AI implementation can get stuck. A weak process means you can’t clearly state what’s supposed to get automated. Poor data means AI is reading garbage from day one, and the result is only as good as the input.

No business owner means nobody on the company side can make a call when an edge case shows up — and edge cases always show up. No maintenance budget means a project that works at launch stops working a few months later, once a document format or a supplier’s system changes. No clear success criteria means nobody will be able to say whether the implementation actually worked.

This isn’t a random list — it’s the five places we most often see implementations get stuck before a single line of integration is even written. We cover the full lifecycle of a project like this, from audit to maintenance, in AI implementation in business: from process audit to ongoing support.

How to read your result

Don’t treat this like an exam with a fixed passing score. If most of your “process” and “data” answers are yes, you have a good pilot candidate — even if the rest of the checklist looks weaker, since some of those items naturally get sorted out during the process audit itself.

If, on the other hand, most of your answers under “business owner” or “maintenance budget” are no, it’s worth pausing before talking specifics. Without an accountable person on the company side and a maintenance budget, even a perfectly chosen process won’t survive much past the first significant change in the input data.

We write about what happens when these elements get skipped in 5 mistakes that keep AI pilots from reaching production — in practice, almost all of them trace back to a gap in one of these five areas.

What to do when the checklist shows gaps

Gaps in the checklist don’t mean you should give up on AI — they show you where to start before committing to a specific implementation. Sometimes it’s one conversation with a decision-maker about who owns the outcome. Sometimes it’s cleaning up data before AI starts reading it. A proper AI-readiness process audit looks at exactly these same five areas — and ends with a concrete recommendation on where to start, not a vague opinion that “AI has potential.”

Frequently asked questions

Do we need to check every box on this list before starting an implementation?

No. The checklist shows where the gaps are, it doesn’t set a passing threshold. Some gaps, especially in the data area, normally get resolved during the audit itself.

What if the company doesn’t have a designated business owner yet?

That’s itself a meaningful result from the checklist — it’s worth appointing that person before starting a conversation about a specific process, since decisions during implementation are hard without one.

Is this checklist only relevant for large companies?

No — we check the same five areas regardless of company size. What changes is the scale of the process, not the list of questions.

How many times should we go through this checklist?

Separately for each process you’re considering for automation — one process might be ready while another at the same company needs more preparation.

Does a low score mean AI won’t work for us?

No — it means that particular process isn’t ready right now. Once the gaps are addressed, for example by cleaning up data or appointing an owner, the same process can become a strong pilot candidate.