There are two ways to spend money on AI in a small business. One is a big bet on a system transformation no one asked for. The other is automating the specific, repetitive work that quietly costs your team hours every week. The second one is where almost every worthwhile AI project in a small business actually starts.
01 The selection test
What is worth automating first
The candidate tasks share three traits: they are repetitive, they follow rules, and they happen at volume. Examples we actually see: routing enquiries to the right place, drafting quotes from a price list, pulling data from one system into another, summarizing follow-up notes, generating standard documents. If a task needs judgement, empathy, or a relationship, it is not a shortcut, and we should stop pretending AI makes that work moot.
- Repetitive. The same shape of task at least a few times a week.
- Rule-based. A clear, describable way to do it correctly.
- Volume. The hours add up, even if each instance is only minutes.
- Data is available. The task runs on data you already have and can access.
02 The pilot
One pilot, chosen for learning, not for glory
Pick one task and automate it properly, end to end, with a human checkpoint where the stakes are low. The pilot exists to answer two questions: does the automation save a real amount of time, and does your team trust the output. Most pilots fail for trust reasons, the output is right but nobody trusted it, so the design has to include the people doing the task, not be built around them.
03 The boundary
Where AI genuinely helps, and where it does not
AI is useful for translation, summarisation, drafting, classification, and pattern-spotting across unstructured data. It is not a reliable source of truth, not something you should trust with decisions by itself, and not a substitute for doing the basic governance around your data. The automation we build keeps a human in the loop at the points that matter, especially wherever money, access, or customers are involved.
04 Where we start
The practical first move
We start with a conversation about your process, not a demo of a tool. You name the one task that eats the most time, and we map whether and how it can be automated, what it would cost, and what a pilot would look like. AI assistants grounded in your own data, integrations between the tools you already run, and workflow automation are all part of that work, and they live on the AI & Automation page. The same repetitive work is what CRM and business systems work clears out, when the fix is a better system rather than more automation.
