AI
AI readiness checklist for SMEs: 12 questions to answer before you spend
Most AI projects that stall were never ready to start. Tick the statements that are true for your business today and you will know whether to run a pilot, commission an assessment or do some groundwork first.
By Wendel Fernandes de LanaFounder, MagzPublished 5 min readIn short
- Readiness comes down to four things: a problem worth solving, usable data, people willing to change how they work, and a way to manage risk and measure results.
- Score 10 or more and you are ready for a focused pilot. Score 6 to 9 and an assessment will show you where to start. Below 6, fix the groundwork first.
- The best first use case is usually a frequent, text heavy task with a clear owner, not the most impressive idea on the list.
On this page
How to use the checklist
Go through the statements below with one or two colleagues who know how the work really gets done, not only the leadership view. Tick a statement only if it is true today, not if it is planned. Be strict: a half true answer is a no.
It takes about ten minutes. Your score updates as you tick, and you can print the page or save it as a PDF to share with your team.
Tick each statement that is true for your business today.
Your score
0 out of 12
0 to 5
Start with groundwork
AI would struggle here today. Write down your key processes, move important information into shared digital systems and agree simple rules for AI tools. These steps pay off even without AI.
6 to 9
Ready for an assessment
You have the basics, but gaps in data, process or risk could stall a pilot. A short assessment will rank your use cases and show which gaps to close first.
10 to 12
Ready for a pilot
You have a clear problem, usable data and people ready to test. Pick the use case with the best balance of value and effort, agree a success measure and run a pilot with real users for four to eight weeks.
What the four areas tell you
Value
AI is a means, not a goal. If you cannot name the task and the number you want to move, any tool will look promising in a demo and disappointing in daily use. A named owner matters just as much: pilots without one fade out when the first problem appears.
Data and knowledge
Language models are good at reading messy text, but they cannot use information they cannot reach. The question is not whether your data is perfect. It is whether the documents, records and emails the task depends on exist in digital form, and whether you can connect them safely.
People and process
Most of the value from AI comes from changing how a task is done, and that depends on the people doing it. A written process, even a rough one, shows where AI fits and where a person should stay in control. Someone also needs time to review the output during testing, because early versions always need correcting.
Risk and measurement
Privacy law applies to AI tools the same way it applies to any other software that handles personal data. Staff are often already using AI tools informally, so a short policy on approved tools and data protects the business quickly. Measuring the task before you start is the only way to show that the pilot worked.
Where SMEs usually start
Good first use cases share three traits: they happen often, they involve reading or writing text, and a mistake is easy to catch before it reaches a customer. Common examples:
- Sorting and drafting replies to customer emails and enquiries, with a person approving before anything is sent.
- Reading documents such as invoices, forms and contracts, and entering the data into your systems.
- Answering internal questions from your own policies, manuals and past projects.
- Summarising meetings, calls and long threads into decisions and actions.
- Preparing reports that someone currently rebuilds by hand every week.
If one of these matches the task you named in the first statement, you already have a strong candidate. Our AI automation service takes tasks like these from manual to automated, with a person in the loop where it matters.
Five mistakes to avoid
- Starting with a tool. Buying licences before choosing the problem leads to a solution looking for a use.
- Picking the hardest use case first. An ambitious first project that fails makes the next one harder to approve.
- Skipping the baseline. Without a before measurement, nobody can say whether the pilot paid off.
- Ignoring the people doing the work. They know the exceptions that break a naive automation, and they decide whether it gets used.
- Leaving privacy until the end. Checking where data goes after the build is slower and more expensive than choosing the right provider at the start.
A quick rule of thumb
If a task happens every day, involves reading or writing, and a person can check the result in under a minute, it is a good first candidate for AI.
What to do with your score
Whatever you scored, the statements you left unticked are your to do list, and most can be fixed in weeks rather than months. If you would like an independent view, a Magz AI readiness assessment reviews your processes, data and systems, ranks your use cases by value and effort, and ends with a written plan for the first pilot.
Want a second opinion on your score?
Message us your score and the task you named in the first statement, and we will suggest a sensible first step.
About the author
Wendel Fernandes de Lana
Founder, Magz
Wendel Fernandes de Lana is the founder of Magz. He holds a Bachelor of Information Systems from the University of São Paulo, completed a summer executive course in digital business at California State University, Northridge, and has worked on web development, CRM automation and Apache AGE, the open source graph extension for PostgreSQL.