What can a small business automate?
Concrete examples of what a smaller business can automate, what suits rules and what suits AI, and where a person still has to look at the work.
By Daniel Aideborn · Published 17 September 2026 · 4 min read
Short answer: whatever happens often, looks roughly the same every time, and is done today by someone moving information from one place to another. Enquiries, reports, documents and syncs between systems are the four that come back in almost every small business.
Below is where the work usually hides, the difference between rules and AI, what you need to contribute, and where the line to human judgement runs.
Four places where there is nearly always work to remove
Enquiries that have to go somewhere
An enquiry arrives by form, email or phone and has to become a case, a row in a list and a reply. What gets automated is the registration, the confirmation to the customer and the reminder to you when nobody has answered. You still write the reply.
Reports rebuilt every week
Someone exports from a system, pastes into a spreadsheet, formats it and sends it. It is the same sequence every week, which is exactly what a machine is good at. The report becomes a by-product of the numbers already being in one place, instead of a task of its own.
Documents typed in by hand
Invoices, delivery notes, agreements and forms that arrive as PDFs or images and have to become fields in a system. AI helps here, because the documents look different every time. The combination that tends to work is a machine that reads and proposes, and a person who approves whatever is uncertain.
Systems that do not talk to each other
A booking system, a till, the bookkeeping, email and a spreadsheet that each work on their own but know nothing about each other. What gets built is a connection that moves the right records in the right direction, with a log of what moved and what failed.
Rules or AI: two different tools
Rules suit work where the answer can be written down in advance: if the field is empty, do this. They are cheap to build, easy to test and never surprising. AI suits work where the input varies: free text, documents in different shapes, summaries. AI costs more to verify, because the answer is not identical every time.
| Task | Rules | AI |
|---|---|---|
| Move an order into the bookkeeping when the format is fixed | Good fit | Not needed |
| Read amounts and dates out of incoming invoices | Rarely enough | Good fit, with review |
| Apply a discount from a price list | Good fit | Not needed |
| Sort incoming free text into the right case type | Rarely enough | Good fit, with review |
| Summarise a long email thread into one note | Not a fit | Good fit |
Most solutions end up a mixture: rules for what is predictable, AI for what is messy, and a person for whatever is expensive to get wrong.
What you need to contribute
- Access to the systems being connected: accounts, permissions and API keys where they exist.
- Real examples, including the awkward ones. Twenty actual cases say more about what is required than a description of how it usually looks.
- A decision about what happens when the automation is unsure: stop and ask, or guess and flag.
- One person who owns the process and can answer questions while it is built.
Where a person still has to look
- Decisions that cost money or affect a customer directly.
- Cases where the input is unclear, incomplete or contradicts itself.
- Personal data in sensitive contexts.
- The first weeks after go-live, however good the test looked.
When it does not pay off
- The work happens a few times a year. Doing it by hand is cheaper.
- The process keeps changing. Automate once it has settled, or you rebuild it every month.
- Nobody owns the process. An automation without an owner becomes a black box no one dares touch.
- The input is too poor. Rubbish in gives rubbish out, faster than before.
If you want to know what can be taken out of your case, the services are described on the home page and the contact box takes a description of the flow. If the question is rather whether the work needs a tool of its own, When is a website enough, and when do you need a web app? is closer to it.
Related services: AI automation