Your business is ready for AI agents when three things are true. The work you want help with follows rules someone could write down, the information the agent needs is written down somewhere it can reach, and there's a clear person who checks what it does. If any of those is missing, an AI agent will either sit there unused or make mistakes faster than your team can catch them.
The ten questions below test those three things. Answer yes or no, and be honest. It takes about five minutes.
The process questions
1. Can you describe the task in steps someone new could follow? If the only way to explain it is "you just know", the agent won't know either, so write the steps down before anything else.
2. Does the task happen at least weekly? Agents pay for themselves on work that repeats. A task that comes up twice a year is hardly ever worth setting up.
3. Is it clear what a good result looks like? A well-qualified enquiry, a record filled in properly, a reply that actually answers the question. If your team argues about what "good" means, the agent can't hit it.
4. Do you know which cases should always go to a person? Complaints, refunds, anything regulated, anything unusual. Knowing the exceptions is what makes it safe to automate the rest.
The information questions
5. Is the information the agent needs written down? Prices, services, policies, processes, the questions people always ask. An agent can only answer from what it's been given.
6. Is that information up to date, and is there one version of it? Say you've got three copies of the price list and each one is slightly different. You'll get three different answers.
7. Does your data live in systems the agent can connect to? A CRM, a calendar, a shared drive. Notes in personal inboxes and paper files are out of reach.
The people questions
8. Is there a named person who will check what the agent does? Especially in the first few weeks. Someone needs to read the conversations, spot the mistakes and change the rules.
9. Will your team trust it enough to use it? If the team works around it, it won't save any time. Get the people who do the work today involved in setting it up.
10. Do you know how you'll measure if it helped? Response time, hours saved, bookings made, errors caught. Pick one number before you start.
What your score means
8 to 10 yes answers: you're ready. Start with the task that scored best on questions 1 to 4, run it with close checking for a month, and then widen it out.
5 to 7: you're close. The no answers are your to-do list, and most of them are about writing things down, not about technology. Fixing them helps your team even if you never use AI.
4 or fewer: start with the groundwork. Write the process down, clean up the information and give it an owner, and then come back to this list. If you bought an agent now, what you'd mostly be buying is a new way to make mistakes.
Where businesses usually start
The first agents that work well tend to be the ones closest to a script.
- Answering and qualifying new enquiries by chat, text or phone, and then booking the right people into a call. See our pages on chatbots and voice agents.
- Answering your team's internal questions from your own documents, with the source quoted. See AI knowledge assistants.
- Turning meeting notes into tasks and draft follow-ups for a person to check before anything gets sent.
What they all have in common is clear rules, information that's written down, and a person who checks the output. That's the whole test in one line.
Take the longer version
Our AI Scorecard goes deeper. It scores your business across each layer an AI system depends on and shows you which one is weakest. It takes a few minutes and you get your results straight away.
