Digital Transformation
Automate three things properly, not thirty things badly
The constraint in an SME is rarely technology. It is that nobody has agreed what a number means.
In short
The reliable sequence for automating a small business is: agree one definition and one source for every number that matters, then automate the two or three workflows that consume the most hours and generate the most errors, then document them so they survive the person who built them. Tooling chosen before that sequence is why most SME technology spend disappoints.
Sequence
What to automate first
| Candidate | Automate now? | Why |
|---|---|---|
| Invoice creation and dispatch | Yes | High volume, rules-based, and the e-invoicing mandate is coming anyway |
| Expense and approval routing | Yes | Removes the single biggest source of month-end delay |
| Payroll and leave records | Yes | Rules-based, high error cost, feeds the gratuity calculation |
| Management reporting pack | Yes, after the data layer | Worthless until each metric has one definition and one source |
| Sales pipeline | Only if the process exists | Automating an undefined sales process encodes the confusion |
| Customer service responses | Partly | First-line drafting yes; judgement and escalation no |
| Inventory forecasting | Rarely, at SME scale | Usually the data history is too thin to beat an experienced buyer |
| Anything a person does twice a month | No | The payback will never arrive |
Data layer
The metric dictionary — the step everyone skips
Before any tool is chosen, write down every number the business manages by: what it means precisely, which system is its single source, who owns it, and how often it is produced. It is usually a one-page table and it usually takes two arguments to complete, because two people will discover they have been using the same word for different things.
That argument is the deliverable. Automation on top of an unresolved definition just produces the wrong number faster and with more confidence.
FAQ
Frequently asked
What is a realistic first-year saving?
We baseline hours and error rates on the specific processes before starting, and measure against that. We do not publish a percentage, because a saving figure quoted before seeing your processes is marketing rather than an estimate.
Should we buy an all-in-one platform or best-of-breed tools?
At SME scale, usually the fewest tools that cover the work, chosen for how well they export data. The integration burden of best-of-breed is consistently underestimated by everyone except the person who has to maintain it.
Do we need a full-time IT person?
Usually not below about 100 people. What you need is a named owner for each system and a documented way to get help — which is a different and much cheaper thing.
Where does AI fit?
After the data layer, not before it. AI applied to inconsistent data produces confident, inconsistent answers, which is worse than the spreadsheet you started with.
Next step
Book a 30-minute briefing
No deck, no pitch. Bring one decision you are stuck on and we will work it through.