The short answer
The best first automation is rarely the most impressive one. It is usually a workflow that:
- happens often (daily or weekly, not twice a year),
- follows rules you can write down without "it depends" on every step,
- starts from digital, structured data (a form, a system record, a standard file), not from phone calls or handwritten notes,
- causes visible problems when it goes wrong, such as late invoices, missed approvals or duplicate records, and
- has an owner who can say what "correct" looks like.
Score each candidate on value and feasibility, start with the ones that score well on both, and measure the result against a baseline you recorded yourself.
Start with an inventory, not a tool
Before looking at automation platforms, list the repetitive work your teams actually do. A short conversation with each team lead usually produces a useful list within a week. Useful prompts:
- What do you copy from one system into another?
- What do you check by hand before something can go out the door?
- What waits in someone's inbox for approval?
- Which reports do you assemble from several sources every week or month?
- What happens every time someone joins, moves roles or leaves?
Write each item as a business event with a start and an end, for example "supplier invoice received to invoice posted in accounting". Vague items such as "finance admin" cannot be scored.
What makes a workflow a good first candidate
Value signals
- Frequency. A task done 40 times a week gives automation far more chances to pay off than one done quarterly.
- Effort per occurrence. Minutes of focused work each time, not seconds.
- Cost of errors. Mistakes that reach customers, suppliers, payroll or the accounts matter more than internal inconvenience.
- Delay cost. Work that holds up something else, such as an order waiting on credit approval.
Feasibility signals
- Rule clarity. You can describe the decision logic, including the exceptions.
- Data readiness. Inputs are already in a system or a consistent file format.
- System access. The systems involved offer an API, a connector or at least a reliable import. Many vendors publish API request limits that vary by plan, as HubSpot does, so check your edition early.
- Process stability. The process is not about to be redesigned or replaced.
Warning signs
- Every case is "special".
- Nobody agrees who owns the process.
- The workflow exists mainly to work around a broken upstream system. Fix or replace that system first.
The prioritization matrix
Score each candidate from 1 (low) to 5 (high). For risk if it fails, a high score means the consequences of a silent failure are serious, so it lowers the priority for a first project. Add your own rows and adjust the criteria if something matters more in your organization.
| 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 written as a business event with a clear start and end.
- Scores were agreed with the people who do the work, not guessed by one person.
- Each row has a named process owner.
- A baseline was recorded for the top candidates (volume, time per case, error or rework count) from your own logs or a two-week tally.
How a filled-in row 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
Plot or sort the results into four groups:
- High value, high feasibility: your first projects. Scope one, deliver it, then use what you learned for the next.
- High value, low feasibility: worth doing, but fix the blocker first (clean the data, agree the rules, get API access).
- Low value, high feasibility: easy wins that can wait, or that a built-in feature may already cover.
- Low value, low feasibility: leave these manual.
Treat a high risk if it fails score as a reason to add review steps, not necessarily to avoid the workflow. Payments and payroll are high-risk, which is why they need approvals, logs and alerts whether or not they are automated.
Common first candidates
These patterns come up in many organizations. Whether they suit yours depends on your scores.
- Re-keying between systems: web orders, form submissions or won deals typed again into the ERP or accounting system. See duplicate data entry and integration.
- Approval routing: purchase requests, expense claims, discounts and credit notes. Many business systems include approval workflows; ERPNext, for example, lets you set multiple levels of approval by role.
- Record set-up: creating a new customer, supplier or item consistently across several systems.
- Joiners, movers and leavers: creating and removing accounts and access when people start, change roles or leave.
- Recurring reports: assembling the same figures from the same sources every week.
Measure before and after
We do not publish typical savings figures because they depend entirely on your volumes, your current process and how well the automation handles exceptions. Instead, record a baseline before you start:
- volume per week or month,
- average handling time per case (a two-week tally is enough),
- errors, rework or complaints linked to the workflow,
- elapsed time from start to finish.
Measure the same things a month after go-live. Include the new work automation creates: reviewing exceptions, maintaining the automation and handling failures.
Limitations and cautions
- Automation hardens a process. If the current process is wrong, automating it makes the wrong outcome faster. Simplify first.
- Exceptions need a home. Decide where records go when the rules do not fit, and who reviews them.
- Someone must own it. Every automation needs an owner, monitoring and a manual fallback.
- Personal information still follows the rules. Moving personal information between systems should stay consistent with what you told people when you collected it, and with the state privacy laws and sector rules that apply to you (for example, the CCPA/CPRA, the Texas Data Privacy and Security Act, HIPAA or the GLBA Safeguards Rule). The FTC's privacy and security guidance is a useful starting point. This is general information, not legal advice.
- AI is not a shortcut past the basics. AI features can help with unstructured inputs such as emails and documents, but the same questions about rules, data and ownership apply. See AI solutions and intelligent automation.
Next step
Fill in the matrix for five to ten workflows. If you want help turning the top one or two into a scoped project, see workflow and business process automation, or use the integration readiness checklist if the work involves connecting systems.
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
- Privacy and security, Federal Trade Commission
This article is general information, not legal, accounting or security advice for your specific situation. Examples are hypothetical unless stated otherwise.