The short answer
The best first automation is rarely the most impressive. Pick a workflow that happens often (daily or weekly, not twice a year), follows rules you can write down, starts from digital, structured data, causes visible problems when it goes wrong (late invoices, missed approvals, duplicate records) and has an owner who knows what "correct" is. Score value and feasibility, start where both are high, and measure against your own baseline.
Start with an inventory, not a tool
List the repetitive work first: a short talk with each team lead usually yields a list within a week. Ask what they:
- Copy between systems
- Check by hand before it goes out
- Leave in an inbox or WhatsApp chat awaiting approval
- Assemble from several sources each week or month
- Do when someone joins, moves or leaves
Write each as a business event with a start and end ("supplier invoice received to invoice posted in accounting"); "finance admin" cannot be scored.
What makes a workflow a good first candidate
| Signal | What to look for | |
|---|---|---|
| Value | Frequency | 40 times a week pays off far more than quarterly |
| Value | Effort each time | Minutes of focused work, not seconds |
| Value | Error cost | Mistakes reaching customers, suppliers, payroll or the accounts |
| Value | Delay cost | Holds something up, like an order awaiting credit approval |
| Feasibility | Rule clarity | Decision logic you can describe, exceptions included |
| Feasibility | Data readiness | Already in a system or consistent file, not calls or handwritten notes |
| Feasibility | System access | API, connector or reliable import; limits vary by plan (HubSpot's), so check your edition; factor the exchange rate into US dollar plans |
| Feasibility | Process stability | Not about to be redesigned or replaced |
The prioritisation matrix
Score 1 (low) to 5 (high). A high risk if it fails means a silent failure would be serious, lowering priority for a first project. Add rows and adjust criteria as needed.
| Workflow (business event) | Frequency | Effort each time | Error cost | Rule clarity | Data readiness | System access | Risk if it fails | Value (F+E+Er) | Feasibility (R+D+S) | Notes |
|---|---|---|---|---|---|---|---|---|---|---|
- Every row is a business event with a clear start and end.
- Scores agreed with the people doing the work, not guessed alone.
- Each row has a named process owner.
- Baseline recorded for top candidates from your logs or a two-week tally.
Example: a filled-in matrix
How filled-in rows might look for a hypothetical distributor:
| Workflow | F | E | Er | R | D | S | Risk | Value | Feasibility |
|---|---|---|---|---|---|---|---|---|---|
| Web order to sales order in ERP | 5 | 3 | 4 | 4 | 5 | 4 | 3 | 12 | 13 |
| Month-end commission report | 2 | 4 | 3 | 3 | 2 | 2 | 2 | 9 | 7 |
| New employee account setup | 2 | 3 | 4 | 5 | 3 | 4 | 4 | 9 | 12 |
The first row is a strong first candidate. The commission report is worth doing later, once the data behind it is cleaner.
Reading your scores
| High feasibility | Low feasibility | |
|---|---|---|
| High value | First projects: scope one, deliver, learn, repeat | Fix the blocker first: clean data, agreed rules, API access |
| Low value | Easy wins that can wait, or a built-in feature may cover | Leave manual |
High risk is a reason to add review steps, not necessarily to avoid the workflow: payments and payroll need approvals, logs and alerts whether automated or not.
Common first candidates
Common patterns; your scores decide whether they suit you.
Re-keying between systems
Web orders, forms or won deals retyped into ERP or accounting: duplicate data entry.
Approval routing
Purchases, expenses, discounts, credit notes; ERPNext supports multi-level approval by role.
Record set-up
New customers, suppliers or items, consistent across systems.
Joiners, movers, leavers
Accounts and access as people start, change roles or leave.
Recurring reports
The same figures from the same sources, every week.
Measure before and after
We do not publish typical savings: they depend on your volumes, process and how well exceptions are handled. Record a baseline:
- Volume per week or month
- Handling time per case (two-week tally)
- Errors, rework, complaints
- Elapsed time end to end
Measure again a month after go-live, counting the new work: exception reviews, maintenance, failures.
Limitations and cautions
Automation hardens a process
A wrong process just produces the wrong outcome faster. Simplify first.
Exceptions need a home
Decide where non-fitting records go and who reviews them.
Someone must own it
An owner, monitoring and a manual fallback, including for when a link or the power is down.
Personal data rules
Moving it should fit the lawful basis and purposes set at collection, a core NDPA principle enforced by the NDPC. Not legal advice.
AI is no shortcut
It helps with emails and documents, but rules, data and ownership still apply (AI automation).
Next step
Fill in the matrix for five to ten workflows. To turn the top one or two into a scoped project, see workflow and business process automation; if systems must connect, use the integration readiness checklist.
Sources and further reading
Product capabilities and guidance change. These are the primary sources this article relies on, checked on the review date above.
- Workflows, ERPNext documentation (Frappe)
- API usage guidelines and limits, HubSpot Developers
- Nigeria Data Protection Act 2023 and implementation guidance, Nigeria Data Protection Commission
This article is general information, not legal, accounting or security advice for your specific situation. Examples are hypothetical unless stated otherwise.