Automating a workflow is not worth it when it saves less than about 5 hours a month, changes every few weeks, or needs a judgement call in more than 1 in 5 cases. Yours is ready when it is stable, measured, and frequent enough that the build pays for itself within 12 months.
When is it not worth automating a workflow, and how do I know if mine is ready?
The break-even math
The test is simple: payback in months = build cost ÷ (hours saved per month × loaded hourly cost − monthly upkeep). If the answer is under 12, build it. Between 12 and 24, use a cheaper tool or wait. Over 24, leave it manual.
Hours saved means minutes per run times runs per month, counting only the part the automation actually removes. Loaded hourly cost is what that person’s hour costs the business. For operations and finance staff, a general market range is $35–$75 once benefits and overhead are included. The examples below use $50.
For the build side, these are Smplfy’s figures. We charge $85 an hour. A typical two-system integration costs $1,500–$6,000, and internal tools start at $4,000. For upkeep we plan on about 20% of the build cost per year. That covers API changes, re-authentication and the occasional fix.
| Workflow | Manual hours/month | Value at $50/hour | Build cost | Upkeep/month | Payback |
|---|---|---|---|---|---|
| HubSpot closed-won to QuickBooks invoice, 60 deals × 20 min | 20 | $1,000 | $3,500 | $60 | About 4 months |
| Same workflow, but 35% of deals need manual review | 13 saved | $650 | $5,000 | $85 | About 9 months |
| Weekly report copied from Stripe into Google Sheets | 4.3 | $215 | $1,500 | $25 | About 8 months |
| Quarterly vendor reconciliation, 6 hours a quarter | 2 | $100 | $2,500 | $40 | About 42 months |
The second row shows how exceptions hurt twice. They cut the hours saved, and they push up the build cost, because every exception needs a rule or a review queue. The last row is a process to leave alone: a checklist will save more than any integration.
The volume threshold
Our rule of thumb, from scoping a lot of these, works by hours of manual work per month:
- Under 5 hours a month: do not hire anyone. Write the steps down, make a template, and look again in six months.
- 5 to 20 hours a month: a no-code tool such as Zapier, Make or n8n is often the right answer, if the steps run in a straight line, the stakes are low, and someone in-house can fix it when it breaks.
- Over 20 hours a month, or anything that touches money or customer records: a custom integration with retries, logging and a runbook usually pays back inside a year.
Frequency matters as well as hours. At fewer than about 20 runs a month, a person catches odd cases more cheaply than code does. Errors avoided, such as a wrong invoice or a duplicate customer, count toward the value only if you can put a dollar figure on them from last quarter.
Signs a process is too unstable or exception-heavy
- The steps changed in the last 90 days, or one of the systems involved is being replaced within six months.
- More than 1 in 5 cases need someone to make a judgement call, and nobody can write down how they decide.
- Two people on the team do the job differently and both think theirs is correct.
- Input arrives as free-form email, phone calls or PDFs in a dozen layouts, at low volume.
- The source data is dirty: duplicate contacts in HubSpot, deals with no customer ID, or amounts typed as text.
- Nobody owns the process, so nobody will notice when the automation quietly stops.
Automating any of these just makes the mess run faster. Exception-heavy work can still be partly automated, though. Handle the common path in code and send everything else to a review queue that a person clears. That split is often the right first build, and AI extraction with a human checking flagged items fits the same pattern once volume justifies it.
A seven-point readiness check
- You can write the steps on one page, and a new hire could follow them.
- The steps have not changed in three months.
- You have measured volume for two weeks: how many runs, and how many minutes each. Not a guess.
- Exceptions are under 20% and you can list them by type.
- Each record has an identifier that matches across systems, such as an email address, deal ID or customer number.
- Every system involved has an API, and you have admin access to it.
- One named person will answer questions weekly during the build and own the workflow afterwards.
If all seven are yes, you are ready to scope it. If only 3 or 4 fail, spend a month fixing the process and measuring it. That costs nothing. If 1, 2 or 5 fail, don’t hire anyone yet.
What breaks after launch
Part of the upkeep line in the table is paying for failures most buyers do not expect. Stored credentials expire, so a connection needs re-authenticating. Someone renames a field in Salesforce. A bulk import hits the API rate limit. A retry creates a duplicate record. A Stripe webhook arrives before the customer record it belongs to exists.
The decision rule
| Your situation | What to do |
|---|---|
| Under 5 hours a month, or payback over 24 months | Leave it manual and write a checklist |
| Steps changed recently or exceptions over 20% | Standardise first, then measure for a month |
| 5–20 hours a month, linear steps, low stakes | No-code tool, maintained in-house |
| Over 20 hours a month, money or customer records, three or more systems | Custom integration with logging and a runbook |
Frequently asked questions
How many hours a month does a workflow need to save to be worth automating?
Our rule of thumb is at least 5 hours a month before automating anything, and about 20 hours a month before a custom build makes sense. Below 5 hours, a written checklist usually saves more than code does. Between 5 and 20, a no-code tool is often enough.
How do I calculate the payback period for an automation?
Divide the build cost by the monthly savings, which are hours saved times loaded hourly cost, minus monthly upkeep. For example, a $3,500 build that saves 20 hours a month at $50 an hour, with $60 a month upkeep, pays back in about 4 months. Aim for under 12 months.
Should I automate a process that has a lot of exceptions?
Not all of it. If more than about 1 in 5 cases need a judgement call, automate the common path and send the rest to a review queue that a person clears. Trying to code every exception usually doubles the build cost and the upkeep.
How much does a small workflow automation cost?
At Smplfy we charge $85 an hour, and a typical two-system integration costs $1,500 to $6,000. Where it lands depends on how many fields need mapping rules and how much error handling the workflow needs. Plan on roughly 20% of the build cost per year for upkeep.
What should I fix before automating a workflow?
Make the steps consistent across the team, clean up duplicate and unmatched records, and measure two weeks of real volume. Make sure every record has an ID that matches across systems and that one person owns the process. If the process is still changing, wait until it has held steady for three months.
Related answers
Not sure your workflow clears the bar?
Send us the workflow and we will reply with a free written scope or a 20-minute call. We will run the break-even math with you, and if it does not pay back we will say so. Have ready which systems are involved, roughly how many records a month, and what is manual today. Start a conversation.