The AI readiness checklist for small businesses
Before investing in AI, check these 8 things. A practical readiness checklist covering data, processes, team, and budget for small business owners.
You've identified a process that's eating up time. You've read about AI automation. You're thinking about investing. Before you do, run through this checklist. It'll save you money and frustration by highlighting gaps that would cause a project to fail.
Not every business is ready for AI, and that's fine. Sometimes the right answer is "fix your data first" or "use simpler automation" or "hire another person." This checklist helps you figure out where you actually stand.
1. Do you have a clear, specific problem?
"I want to use AI" is not a problem statement. "My accounts team spends 15 hours a week manually matching invoices to purchase orders" is.
The problem needs to be:
- Specific: what task, done by whom, how often?
- Measurable: how many hours, what error rate, what cost?
- Repetitive: does it happen frequently enough to justify automation?
If you can't describe the problem in one sentence with a number in it, you're not ready. Go back to the time audit and identify the bottleneck first.
2. Is your data digital and accessible?
AI needs data. If the information relevant to your problem is trapped in:
- Paper files or filing cabinets
- People's heads ("only Dave knows how to...")
- Email threads with no structure
- A legacy system with no API or export function
Then you have a data infrastructure problem, not an AI problem. Fix the data access first. This might mean digitising paper records, building a simple database, or migrating to modern tools that have APIs.
Note
A data cleanup project might sound like wasted money, but it pays off regardless of whether you build AI on top of it. Accessible, structured data makes everything faster: reporting, onboarding new staff, answering management questions, and eventually, automation.
3. Do you have enough data?
Different AI approaches need different amounts of data:
| Approach | Minimum data needed | |---|---| | Rule-based automation | None (just clear logic) | | LLM with good prompting | None (but needs clear instructions) | | Text classifier (fine-tuned) | 200-500 labelled examples | | Predictive model | 1,000+ historical data points | | Custom computer vision | 500+ labelled images |
If your business has been operating for a year or more and tracks its operations digitally, you probably have enough for most approaches. If you started tracking data last month, give it time.
4. Is the data consistent?
Having data is one thing. Having clean, consistent data is another. Common problems:
- Inconsistent labels: "Customer complaint," "Complaint," "COMPLAINT," and "cust. comp." all meaning the same thing
- Missing fields: Half the records are missing key information
- Format changes: The spreadsheet format changed three times in the last two years
- Duplicates: The same record appears multiple times with slight variations
You can clean data during an AI project, but it adds time and cost. If you know your data is messy, budget at least 30-40% of the project time for cleanup.
5. Can the process tolerate errors?
AI is probabilistic. A good classifier might be 95% accurate, which means 1 in 20 predictions is wrong. For some processes, that's fine, especially with human review. For others, it's not acceptable.
Good fit for AI (tolerates errors):
- Support ticket triage (a mis-routed ticket gets re-routed by the agent)
- Lead scoring (a wrongly scored lead still gets reviewed)
- Document classification (a misfiled document gets corrected)
Poor fit for AI (can't tolerate errors):
- Medical diagnoses
- Financial compliance reporting
- Safety-critical decisions
If your process requires 100% accuracy and has no room for human review, AI is the wrong tool. Use rules-based automation instead.
6. Is there someone who can own it?
An AI system needs an internal owner. Someone who:
- Understands the process being automated
- Can review the system's output and flag problems
- Communicates with the developer when something breaks
- Makes decisions about edge cases and business rules
This doesn't have to be a technical person. It's whoever currently manages the process. But they need to be available and engaged, not treated as an afterthought.
7. Do you have a realistic budget?
Rough ranges for AI projects in the UK:
- Simple automation (rules-based): £2,000-5,000
- ML model with integration: £5,000-15,000
- End-to-end system: £10,000-30,000
- Monthly maintenance: £500-2,000
Plus hosting and API costs, which vary by usage but are typically £50-300/month for small business workloads.
If your budget is under £5,000, you can still do a lot with no-code automation and simple scripting. If it's under £2,000, focus on getting your data infrastructure in order: that investment will pay off later.
8. Is the timing right?
Don't start an AI project during:
- A major system migration (switching CRM, ERP, or accounting software)
- A restructuring (the process you're automating might change)
- Peak season (staff won't have time to test and provide feedback)
- Budget uncertainty (a half-finished AI project is worse than no AI project)
The best time is when operations are stable, the team has bandwidth to test the system, and the budget is confirmed for the full project including maintenance.
Key Takeaways
- Start with a specific, measurable problem. 'I want AI' isn't enough.
- Digital, accessible data is the foundation. Fix your data infrastructure before building AI.
- Budget for data cleanup: 30-40% of most AI projects is data preparation.
- The process must tolerate some errors. If it can't, use rules-based automation instead.
- Assign an internal owner. AI systems need someone who understands the business process to manage them.
Want to check your readiness?
If you've run through this list and most boxes are ticked (or you're not sure about a couple), get in touch. I can do a quick assessment over a call and tell you whether you're ready to build or whether there's groundwork to do first.
Related reading: