The businesses that get real value from automation don't start with the technology. They start with one repetitive process worth fixing — and avoid the two mistakes that sink most first projects.
The most common mistake businesses make with AI and automation is starting with the tool — "we should use AI for something" — rather than the problem. The better starting question is much narrower: what repetitive, rules-based task is eating time that a skilled person shouldn't be spending on it? Invoice data entry, appointment scheduling, first-pass customer enquiries, chasing internal approvals — the answer is usually something already familiar, not something exotic.
| Good first candidate | Poor first candidate |
|---|---|
| Happens daily or weekly | Happens once a quarter |
| Rule-based, even with some variation | Judgement-heavy every time |
| Currently done by a person who could add more value elsewhere | Already fast and low-cost to do manually |
| Success is easy to measure | No clear definition of "done well" |
| Process is documented and consistent | Process changes every few weeks |
Processes that are inconsistent, undocumented, or that change constantly are usually the wrong place to start — not because automation can't eventually help, but because you'll be automating confusion rather than fixing it.
Before starting, agree what "working" looks like — hours saved per week, a reduction in manual errors, faster turnaround on a customer-facing process. Without a baseline measured before you start, it's hard to know afterwards whether the automation actually delivered, or just felt like progress.
The businesses that get sustained value from automation treat the first project as a proof point, not a one-off. Once one process is automated cleanly, with a measurable result, it's much easier to identify the next candidate — and to get buy-in for it — than trying to automate everything at once from a standing start.
Once you know what "a good candidate" looks like, the next useful question is what's actually doing the work behind the scenes — see AI virtual agents vs chatbots vs RPA.
For a well-scoped, single workflow, weeks rather than months is realistic — the scoping and process-fixing usually takes longer than the technical build itself.
Document and fix it first. Automating an undocumented, inconsistent process just automates the inconsistency — the documentation step is part of the project, not a delay to it.
No — starting with one well-chosen, narrow process deliberately keeps the first project small. The goal of the first project is proving value, not transforming the whole business at once.
Talk to our team about where automation could make the biggest difference in your business — no commitment required.
