The best first automation is usually the dullest job in the office. It’s done the same way every time, by someone copying information from one system into another, and a mistake costs money or a customer. Tracking numbers onto invoices, supplier price lists into your stores, and payments matched against what’s owed are typical examples.
To find yours, list the jobs your team repeats, time each one, and rank them by how often they happen, how long they take and what an error costs. Then cross off anything that needs judgement, happens rarely or is broken as a process. What’s left at the top of the list is where to start.
Signs a task is worth automating
Most business processes that can be automated share four traits. The more of them a task has, the better a candidate it is.
- It’s repeated the same way. Every order, every Monday, every new client. If two staff members do it the same way without talking to each other, software can probably do it too.
- The rules can be written down. “If the order is paid and the tracking number exists in ShipStation, add it to the invoice in QuickBooks.” If you can write the steps as plain sentences, a developer can turn them into code.
- The data is already in a system. Shopify, Xero, MYOB, Stripe, a supplier’s spreadsheet, a shared inbox. Automating data entry is far easier when the information already exists somewhere and someone is just moving it.
- Errors are costly. A missed payment, a wrong price on a store, a changed size that reaches the factory saw. Software doesn’t get tired at 4 pm on a Friday, and it can check every record, not a sample.
Signs it isn’t, at least not yet
Some work looks repetitive but shouldn’t be handed to software. Saying so early saves you money.
- It’s a judgement call. Deciding whether to extend credit to a struggling customer, or how to word a reply to an upset client. AI can draft or sort, but a person should decide. On the Raw Blend inbox, Claude drafts replies and staff read and send every one.
- It’s rare. A job done twice a year rarely pays back the cost of building and maintaining an automation. A checklist is the better answer.
- The process itself is broken. If three people do the job three different ways, or nobody agrees on the rules, automating it just makes the mess run faster. Fix the process first, then automate the fixed version.
- An app already does it. If Xero or Shopify has a built-in feature or a well-supported add-on that fits your process, use it. Custom work makes sense where nothing off the shelf fits how you actually work.
How to measure a task before you automate it
You don’t need a consultant for the first pass. A notepad and a fortnight is enough. For each repeated job, write down:
- How long it takes each time. Time it with a stopwatch, not from memory. People underestimate small jobs they do often.
- How often it happens. Per day, per order, per week. Multiply by the time to get hours per week.
- Who does it. A job done by the owner or a senior bookkeeper costs more than the same hours from a junior. It also takes them away from work only they can do.
- What goes wrong. How often a mistake happens, and what it costs when it does: a refund, a re-make, a late payment chased for weeks.
- Which systems it touches. Note every program, file and inbox the job passes through. This tells you how hard it will be to automate.
One clinical nutrition practice owner did this for a week and found 28 hours of manual work across 14 tasks, with 7 of those hours spent on money. That list became the scope for the build. Once the numbers are on paper, the ranking usually becomes obvious. A job that takes ten minutes but happens forty times a day beats a two-hour job done once a month.
If you’d rather someone watch the work and map it with you, that’s what the Bottleneck Report does. It ends in a written, ranked list, including the jobs that aren’t worth building.
Common first wins, with real examples
These are the kinds of jobs that tend to top the list, taken from work we’ve built.
Tracking numbers onto invoices
At Raw Blend, warehouse staff copied tracking numbers from ShipStation onto QuickBooks invoices by hand. The first automation adds new tracking numbers to the matching invoices every hour and skips ones already done. On its first run it matched and updated 361 of 383 invoices, clearing a month’s backlog. The other 22 were ambiguous, so they went to a person instead of being guessed. That one sync took away about an hour a day of re-typing. You can read how it grew in the Raw Blend case study.
Supplier price lists
Raw Blend buys from 26 suppliers, and their price lists arrive as spreadsheets, CSV files and portal downloads. Someone used to key the changes into every store. Now the hub reads each list and updates costs in QuickBooks and prices and stock across the stores. It also refuses big swings: if a list would move more than 25% of a range by over 10%, it isn’t applied, because that usually means a broken file. The check matters as much as the automation.
One job, entered once
A Brisbane blinds and shutters manufacturer was typing every job three times: into an old quoting program, into MYOB and into spreadsheets. In the system we built, now in acceptance testing, a job is entered once as a quote and becomes the order, the factory work order and the invoice. Removing double entry is often the biggest single win, because every extra copy is another chance for a size or colour to be wrong.
Payment checks
The clinical nutrition practice took payments through four channels and matched them by hand. Its hub now runs an hourly check for any paid invoice it missed. On the first run, it found two $1,500 deposits that had been paid and missed. A check like this is small to build, and here it found $3,000 on its first run.
Why a central hub beats a pile of separate automations
Your first automation can be small. A single automation starts from about $1,000 and can be running in about a week. The trouble starts with the fifth or sixth, if each one is built in a different place.
A business that adds automations one at a time often ends up with a Zapier zap here, a script on someone’s computer there and a sync tool nobody owns. Each one works on its own. Together, nobody knows what runs when, the same customer exists in three versions, and when a supplier changes a file something fails without telling anyone.
A central hub keeps the records and the rules in one place. Each record is entered once, and every automation reads from and writes to the same source. You get one run history, one place to see what failed and one set of business rules, such as the price-swing limit, applied everywhere. Raw Blend’s hub started with the tracking sync and now runs 52 automations across six Shopify stores and one QuickBooks company. New jobs plug into what’s there instead of starting from scratch.
That doesn’t mean building the hub first. Start with one job that’s clearly worth it, built so the next one can sit beside it.
A practical next step
This week, ask each department to list the three jobs they repeat most. Time each one for a fortnight, note who does it and what goes wrong, and rank them with the signs above. Cross off anything that needs judgement or a process fix. The task at the top of what’s left is your first automation.