The conversation about artificial intelligence in SMEs almost always runs backwards: it starts with the tool ("shall we get ChatGPT for the team?", "shall we add a chatbot?") rather than with the problem. The result is licences nobody uses and demos that never reach production. An AI audit is exactly the opposite: first look at how your company works, then decide where AI adds value and where it doesn't.
What an AI audit is (and isn't)
It isn't a trends report or a list of fashionable tools. It's a process diagnosis with one question behind each process: is there human time here spent on something repetitive, based on text or data, that a machine could do as well or better? And its twin: how much is that time worth? The output is a map of opportunities ranked by return and by ease, not by how impressive they look.
How it's done, step by step
- Process inventory. Short interviews with each area: sales, admin, customer service, operations, marketing. What they do every day, with which tools and where they get stuck.
- Measurement. Weekly hours per task, volume (how many emails, quotes, tickets, product sheets) and the cost of errors. Without numbers there's no priority.
- Data map. Where the information lives: ERP, CRM, email, Excel, Drive. AI without accessible data does nothing; very often the first step isn't AI but tidying up.
- Use cases. For each opportunity: what the AI would do, what risk it carries (sensitive data, costly errors, customer-facing), which tool or development solves it and the estimated return.
- Prioritisation and roadmap. Two or three measurable pilot projects over 90 days. Never ten at once.
What usually comes up in an SME
After auditing businesses in very different sectors, the patterns repeat: replying to emails and enquiries (quotes, questions, bookings) that eats up the hours of qualified people; classifying and extracting documents (supplier invoices, orders, CVs); reports that someone assembles by hand every Monday; repetitive content (product sheets, translations, descriptions); and out-of-hours customer service. It almost never turns out to be "we need our own model". It almost always turns out to be "we need to connect what we already have and automate three things".
The uncomfortable questions you also have to ask
What customer data are you going to run through a model, and with what guarantees? Who reviews what the AI produces before it reaches the customer? What happens the day the tool changes its pricing or disappears? A serious audit covers risk, compliance and dependency, not just opportunities. And an honest write-off too: there are processes where AI isn't worth it, and saying so is part of the job.
Why start here
Because it's cheap compared with getting it wrong. A two- or three-week audit saves months of projects with no owner. At staycreative it's the entry point to our artificial intelligence services: first we understand, then we automate, develop or train, and only what has a return. If you want to know what AI would do for your business before spending a single euro on licences, start with an audit.
