AI readiness checklist: 8 questions to answer before you automate
Updated:
An AI readiness check tells you which work in your company is worth handing to AI, and whether your data and systems can support it. A smaller company doesn’t need a scored maturity model for that. It needs honest answers to the eight questions below, and one process to start with.
What does an AI readiness check cover?
Four things: the work you want to hand over, the data it would run on, the systems it has to connect to, and the people who stay responsible for the result. Fill in the checklist per process, not for the whole company. “Customer service” is too broad to answer. “Copying incoming orders into the invoicing system” is not.
The AI readiness checklist
- Can you name the job? Write down in one sentence who does what and what starts it. If you can’t, the process needs clarifying first.
- Does it mostly run the same way? AI handles varied text and some exceptions well, but it needs a usual path to be measured against.
- How much time does it take each month? Multiply how often it happens by how long it takes. If you don’t know, measure it for two weeks.
- Does the input arrive digitally? Emails, PDFs and forms are a good start. Paper and phone notes need digitising first.
- Can your systems be reached? Check whether the system the data comes from, and the one it goes to, has an API, an integration or at least an automatic export.
- What happens when it’s wrong? AI makes mistakes. Decide where checking afterwards is enough and where a person must approve before anything goes out.
- What personal data would it see? Know where the data is stored and who the provider is. If an AI provider processes personal data on your behalf, you need a data processing agreement (GDPR Article 28). The rest is in GDPR and AI.
- Who owns it afterwards? Name one person who checks the results and speaks up when something breaks.
How do you score the answers?
Don’t add up points. Sort each process into one of three groups:
- Start here: the first five answers are yes, and mistakes are easy to spot.
- Fix first: the work is worth it, but digital input or system access is missing.
- Leave with a person: it happens rarely, takes a lot of judgement, or you wouldn’t catch a mistake in time.
Pick one process from the first group, not several. If every case is exactly the same, you may not need AI at all: a plain rule is cheaper and more predictable.
What comes after the checklist?
Measure how long the process takes today. Run a small pilot on real data, with a person checking the results. Then compare the time and the mistakes, and only move to the next process if the numbers support it. The full sequence, from choosing the process to rolling it out, is in how to implement AI in a business.
Can you do this yourself?
Yes. For one or two processes, this checklist is enough. An outside view helps when many systems are involved, you work with personal data, or nobody in the company has built something like this before. vSites does this as AI consulting: we go through your processes and tell you in writing, with a fixed-price plan, what is worth handing to AI. If it is, we also build the AI automation. If AI isn’t the answer, we say so.