Find the work worth automating
Automation pays where work is frequent, rule-based and error-prone. Spend a week writing down what people retype, chase or check by hand.
For each item, note how often it happens, how long it takes and what a mistake costs. A task done twice a year is seldom worth automating. A task done forty times a day often is.
Common examples
| Manual work | What automation looks like | What to check |
|---|---|---|
| Matching mobile money to invoices | A payment arrives and clears the right invoice. | What happens with a wrong reference or an overpayment? |
| Approving purchases and expenses | Requests go to the right person based on rules, and the decision is recorded. | Can approvals be delegated? Can someone approve their own request? |
| Posting to the ledger | Sales, receipts and stock movements post themselves. | Can the accountant see and correct the mappings? |
| Chasing debts | Reminders and statements go out on a schedule. | Can you stop them for a disputed invoice? |
| Reorders | Low stock triggers an alert or draft order. | Are lead times and minimum levels up to date? |
| Owner reports | A summary arrives each morning. | Do the figures match the full reports? |
Selection criteria
- Fit with your systems. Does the automation live in the system that holds your data, or does it depend on a link between systems?
- Controls. Are there permissions, approval limits and a record of who did what?
- Failure handling. When a step fails, does someone find out, and can it be retried without duplicates?
- Who can change it. Can a trained manager change a rule, or does every change need a developer?
- Integrations and API. Does it connect to your bank, payment, tax and messaging providers? Is there an API, and is it on your plan?
- Running cost. Some charge per task, message or user. Estimate your real volumes.
- Reports. Can you measure whether the automation saved time?
Three kinds of tool
No type is better in every case. Choose by where the data lives.
| Type | Examples from vendor pages | Good when | Watch for |
|---|---|---|---|
| Automation inside your accounting or ERP | Business Central agents and Copilot; Navigate approvals and auto-posting; Odoo apps; Zoho Inventory workflows and webhooks | Most of your work is in one system. | Tied to that vendor. Plan gating. |
| Connected app ecosystems | Xero's app store with over 1,000 connected apps; Zoho's integrations | You use several strong tools. | Each link can fail, and costs add up. |
| Open and customisable platforms | ERPNext; Odoo Studio on the Custom plan | You have developers and specific needs. | You own the upkeep, and custom code may need separate maintenance. |
Be careful with AI claims
Many vendors now describe AI features, for example agents for expenses and sales orders in Business Central and document reading in TallyPrime. These may help. Ask to see them with your own documents, ask what happens when the AI is wrong, and keep a person checking anything that affects money or tax.
What Navigate automates
Navigate publishes this guide and is one option. Its automation is mostly built into its own processes:
- Journals post automatically from sales, receipts, stock and other documents.
- M-Pesa prompts, callbacks and matching to customers. Bank statement lines can be matched automatically once statements are entered or imported.
- Approval workflows, with one place for what waits for each person.
- System controls: your own rules for fields a document must have and checks it must pass before saving or posting.
- Scheduled report emails and an owner's daily report by WhatsApp and email, and low-stock alerts by push notification.
- A versioned REST API with API keys, on plans with API access.
A sensible rollout
- Pick one process and write down its rules on one page.
- Clean the data it depends on.
- Run it for a month alongside a manual check.
- Measure time saved and errors, then decide on the next one.
Where Navigate may not fit
- Navigate has no visual builder for general workflows across outside apps. Consider a dedicated automation tool for that.
- Live bank feeds are planned, not available yet. Statements are entered or imported before matching.
- Direct mobile money automation is for M-Pesa. Other providers go through a card or wallet checkout where supported, or are recorded by hand.
- The REST API is not on every plan.
Frequently asked questions
What should a business automate first?
Start with the most frequent, rule-based task with a clear cost of errors. For many businesses that is posting transactions, matching payments or routing approvals.
Will automation replace our accountant?
It should remove repeated entry, not judgement. Someone still needs to review results, handle exceptions and sign off the numbers.
Do we need a separate automation tool?
Often not at first. Use what your main system already does. Add a separate tool when you must link systems that have no built-in connection.
How do we measure whether automation worked?
Record the time taken and error rate before you start, then measure again after a month. Also track the failures the automation itself creates.