Hiring vs. Automation ROI: How to Compare the Costs
In short
A hire costs about 1.35 to 1.43 times salary. Automation costs a build plus software and review time. Here is the break-even math with a worked example.
Automating a task beats hiring for it when the build cost pays back in a period you can accept. Payback in months is the build cost divided by the monthly savings, and the monthly savings are the loaded cost of the human work removed minus what the automation costs to run, including the time people spend reviewing it. The input that decides most cases is how much of the work the automation actually takes over.
The Loaded Cost of a Hire
A salary is not the cost of an employee. The Bureau of Labor Statistics tracks what employers pay per hour worked, and the ratio of total compensation to wages gives a multiplier you can apply to any salary:
| Employer group (June 2026) | Wages per hour | Total compensation per hour | Multiplier |
|---|---|---|---|
| All private industry | $32.82 | $46.89 | 1.43× |
| Establishments with 1 to 49 workers | $27.88 | $37.64 | 1.35× |
The benefits part covers paid leave, insurance, retirement contributions and legally required costs such as the employer share of Social Security and Medicare. It leaves out recruiting, equipment, software seats, office space and the manager's time, so the true figure is higher.
To complete the hire side, add your own one-time costs (job board fees, interview hours, a laptop) and recurring ones (software seats, desk space). A new hire also takes time to reach full speed, so the first months deliver less than a full salary's worth of work. The automation side has a similar gap while the build is tested.
The Automation Cost Curve
Automation, whether a script, a workflow tool or an AI agent, has four cost lines:
- Build: the one-time cost to design, test and launch it.
- Software and usage fees: subscriptions or API charges, which often rise with volume.
- Maintenance: updates when the process, the data or a connected tool changes.
- Review and exceptions: the hours a person spends checking output and handling cases the automation can't. This is the line most often left at zero, and it is rarely zero.
Worked Example (Hypothetical Role)
The hire: $48,000 salary × 1.4287 (BLS ratio, $46.89 ÷ $32.82) = $68,578 a year loaded.
The automation (assumptions):
- Build: 200 contractor hours × $150 = $30,000, one-time.
- Software and API: $400 a month = $4,800 a year.
- Review: 5 hours a week × 50 weeks × $60 loaded = $15,000 a year.
- Maintenance: 10% of the build = $3,000 a year.
- Running cost: $4,800 + $15,000 + $3,000 = $22,800 a year.
Case A, automation does all the work: savings = $68,578 − $22,800 = $45,778 a year. Payback = $30,000 ÷ ($45,778 ÷ 12) = 7.9 months.
Case B, automation handles 60%: a person still does 40%, costing 0.4 × $68,578 = $27,431. Total = $27,431 + $22,800 = $50,231. Savings = $18,347 a year, and payback stretches to 19.6 months.
The build price is the same in both cases. What changed the answer was the share of the work that moved off a person's plate. Before approving a build, test it on a sample of real cases and count how many it handles without help.
How Payback Changes With Coverage
Keeping every other input from the example, here is how payback moves as the share of the work the automation handles changes:
| Share automated | Total yearly cost | Yearly savings | Payback on $30,000 |
|---|---|---|---|
| 100% | $22,800 | $45,778 | 7.9 months |
| 80% | $36,516 | $32,062 | 11.2 months |
| 60% | $50,231 | $18,347 | 19.6 months |
Total yearly cost is the remaining human share of the $68,578 loaded salary plus the $22,800 running cost. If you plan to redeploy the person rather than not hire at all, the savings only count if the new work they take on is worth at least as much.
Which Tasks Tend to Cross Over
| Task trait | Leans toward automation | Leans toward a hire |
|---|---|---|
| Volume | High and steady | Low or irregular |
| Rules | Written down and stable | Change often or rely on judgment |
| Inputs | Structured forms and fields | Messy emails, calls, documents |
| Cost of an error | Low and easy to catch | High or hard to spot |
| Accountability | None needed per case | A named person must own it |
Partial automation is common. In a study of customer-support agents given a generative AI assistant, published by the National Bureau of Economic Research, issues resolved per hour rose 14% on average and 34% for novice and lower-skilled agents, with little effect for the most experienced. The tool made people faster; it didn't do the job alone. That pattern looks like Case B, not Case A.
What This Doesn't Cover
This guide doesn't cover employment law when a role is eliminated, the tax treatment of software development costs, ramp-up time for a new hire, or the customer impact of automation errors. Talk to an accountant about how build costs are deducted, and to an employment attorney before eliminating a position.
AI Productivity Calculator
Estimate net hours saved after review time.
AI Agents vs. Hiring
The same comparison for AI tools and a growing team.
Sources
Found this useful?
Copy the link to share it.