Is my business ready for AI, and what should I fix before I spend money?
Short answer: Your business is ready for a first AI project if you can answer yes to most of ten plain questions about your process, data, owner and review step. An AI readiness checklist is a list of conditions that make an AI project likely to work, such as a written process, tidy records and one named person in charge. Anything you cannot answer yes to has a fix that usually takes a few hours to a few weeks, and you can do most fixes yourself before spending anything.
You want a straight list of what has to be true in your business before AI is worth paying for, with no form to fill in.
I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.
If you would rather have someone work through this with you, the AI Strategy Sprint does it in 3 days and ends with written steps.
How to use this checklist
There are ten questions below. For each one, answer yes or no for one specific job you want AI to help with, not the whole business. Pick a single job first, such as answering new enquiries or writing quotes. Readiness is about a job, not a company.
Each question has a pass test and a fix. The fix has a time estimate. These are my own rough estimates for a team of 5 to 50 people, not measured figures, so treat them as a guide. Nothing here is gated and there is no score to email you.
| # | Question | Area |
|---|---|---|
| 1 | Can you describe the job in 10 steps? | Process |
| 2 | Does it happen often enough to matter? | Value |
| 3 | Do two people do it the same way? | Process |
| 4 | Are the inputs written down somewhere? | Data |
| 5 | Is that data in one place, readable by software? | Data |
| 6 | Do you know what must never go into an AI tool? | Privacy |
| 7 | Is one person named as the owner? | People |
| 8 | Will a person check the output before it matters? | Review |
| 9 | Does your team want this, or fear it? | People |
| 10 | Can you tell if it worked? | Measure |
Questions 1 to 3: process
1. Can you describe the job in 10 steps?
Pass: you can write the job as a short list: what starts it, what gets checked, what gets sent, who gets told. Fix: sit with the person who does it and write the steps as they happen, not as they should happen. About 1 to 2 hours. If you cannot describe it, AI cannot do it, because an AI tool follows instructions and context you give it. A written process is also the base for a standard operating procedure, and my guide on business process automation shows how to turn it into a flow.
2. Does it happen often enough to matter?
Pass: at least several times a week, or it eats at least a few hours a week in total. Fix: if it is rare, leave it alone. Count it for two weeks first. About 10 minutes a day.
3. Do two people do it the same way?
Pass: a new hire would do it roughly like your best person. Fix: agree on one version before you teach it to software. If your team does it five ways, the AI will learn the mess. About half a day of agreeing, plus a short written note.
Questions 4 to 6: data and privacy
4. Are the inputs written down somewhere?
Pass: the facts the job needs (price lists, past quotes, policies, customer history) exist in a file or system, not only in someone's head. Fix: write down the top 20 questions customers ask and the real answers. About a day. This is the same material that feeds an AI knowledge base.
5. Is that data in one place and readable by software?
Pass: one current copy, in a spreadsheet, a database or a tool with an export. Fix: pick one home for each type of data and retire the old copies. Scans of paper and photos of whiteboards are the hardest to use. About 1 to 3 weeks depending on the mess. Data readiness is where many projects stall, so give this question the most thought.
6. Do you know what must never go into an AI tool?
Pass: you have a short list, such as customer health details, card numbers, passwords and anything under a confidentiality agreement. Fix: write that list and check what each tool does with what you paste. Business plans of AI assistants often say they do not train on your inputs by default. For example, Anthropic, the company that makes Claude, says on its support page that it does not use inputs or outputs from its commercial products, such as Claude for Work and its API, for training by default. Consumer plans can differ, so check each tool's own terms. My guide on keeping your data out of AI training walks through the questions, and my one-page AI usage policy puts the answers in writing. The US FTC's business guide on protecting personal information gives a simple rule: keep only the data you need. About 1 hour for the list, 1 to 2 hours to check your tools. Ask a lawyer if you handle health, financial or children's data.
Questions 7 to 9: people and review
7. Is one person named as the owner?
Pass: a named person is responsible for the result, and it is not "everyone". Fix: pick the person who feels the pain of the job most. They need an hour or two a week. Without an owner, projects fade after the first month.
8. Will a person check the output before it matters?
Pass: anything that reaches a customer, touches money or changes a record is approved by a person at first. Fix: design the check into the flow: the AI drafts, the person approves with one click. AI language models can state wrong things in a confident voice. OWASP, which publishes security guidance for AI applications, lists misinformation and excessive agency, meaning too much freedom to act without oversight, among its top risks. Drop the check only after the draft has been right for weeks. About 1 hour to design.
9. Does your team want this, or fear it?
Pass: the people who do the job have been told what is changing and asked what annoys them. Fix: hold a 20-minute talk. Say what the tool will do, what it will not do, and that it takes the dull part of the job, not the job. People who feel looped in report problems. People who do not will work around the tool. NIST's AI Risk Management Framework, a voluntary guide from the US standards body, puts culture and accountability under its Govern function for the same reason.
Question 10: can you tell if it worked?
10. Do you have a before number?
Pass: you know today's figure for the job: minutes per task, reply time, or how many quotes get followed up. Fix: measure it for one week before you start. About 10 minutes a day. Without a before number, you cannot prove the tool helped, and you may keep paying for something that does not.
Reading your answers
- 8 to 10 yes: you are ready. Pick the one job and start small. Try it with a general AI assistant first. My AI implementation plan shows the steps.
- 5 to 7 yes: fix the data and process gaps first. That is time well spent, because the same fixes help even without AI.
- Fewer than 5: do not buy anything yet. Write the process down, collect the data in one place and name an owner. That is the project.
Do not read this as a score. One no on question 6 or 8 can matter more than five yeses, because those two are about harm, not speed.
A worked example
This example uses made-up details to show the format. A 15-person bookkeeping firm wants AI to draft replies to client emails.
- Q1: yes, the steps are clear.
- Q2: yes, about 40 emails a day.
- Q3: no, three partners answer differently. Fix: one half-day to agree on standard answers.
- Q4 and Q5: no, the answers sit in old emails. Fix: copy the 25 most common answers into one document. Two days.
- Q6: partly, they pasted client tax details into a free chatbot last month. Fix: write the never-paste list and move to a business plan.
- Q7: no owner. Fix: the office manager takes it.
- Q8: yes, partners approve.
- Q9: yes after a short talk.
- Q10: no. Fix: count reply times for a week.
Result: five fixes, about a week of part-time effort, then ready. The money they would have spent on a tool the first month is saved.
When a sprint beats self-assessment
You can do all of this alone. It gets hard when you have several jobs competing for the first slot, when staff disagree on how the work is done, or when you do not know which of your tools already have AI inside them. In those cases an outside view saves time. That is what a short, paid sprint is for, and I cover how to judge one in my guide to what an AI audit includes and costs.
How I'd approach this
I would pick one job, answer the ten questions in an afternoon, and fix only what fails. I would never start with a tool. I would start with the person who does the job and the written steps. Many small firms are closer to ready than they think. The usual missing pieces are an owner, a written never-paste list and a before number, all cheap to fix.
If you want help ranking jobs, the AI Strategy Sprint is a paid 3-day first step where you and I find where AI could transform the business and generate new value, and you get the exact steps written down. It begins with a free 15-minute AI brainstorm call.
Questions people ask
What does AI readiness mean for a small business?
It means the basics are in place so an AI project can work: a described process, usable data, a named owner, a review step and a way to measure results. It has little to do with company size or technical skill.
Do I need clean data before using AI?
You need data that is findable and mostly correct for the one job you pick, not a perfect company-wide system. Start with the 20 questions customers ask most and the true answers. My AI knowledge base guide shows how.
How long does it take to get ready?
For one job in a 5 to 50 person firm, my rough estimate is a few days to a few weeks of part-time effort, depending mostly on how scattered your data is. It is an estimate, not a measured average.
Should my staff be worried about AI replacing them?
Talk about it openly. Tell them which dull task is changing and that people will review the output. Staff who understand the plan tend to help improve it.
Do I need an AI policy before I start?
A short one, yes. It should name approved tools, what must never be pasted in and who reviews output. Use my one-page AI usage policy template.
Is it safe to use AI with customer information?
It can be, with business-grade plans, a clear never-paste list and checking each vendor's data terms. Anything involving health, financial or children's data needs a lawyer's view. This is general information, not legal advice.
More in this topic: Problem solving: AI strategy and adoption (24 guides) →
Sources
- NIST: AI Risk Management Framework
- NIST: AI RMF Playbook
- NIST: AI RMF Core (Govern, Map, Measure, Manage)
- OWASP Gen AI Security Project: Top 10 for LLM Applications
- FTC: Protecting Personal Information, A Guide for Business
- FTC: Data security guidance for business
- Anthropic: Is my data used for model training?