Measuring the ROI of Automation Honestly, Before and After Go-Live
Automation business cases often promise savings that never appear in the accounts. Measuring ROI honestly means counting every cost and being clear about what saved hours are really worth.

Almost every automation proposal comes with a number. Hours saved per month, a payback period, sometimes an impressive percentage. A year later, few organisations check whether the number came true, and when they do, they often find that the hours were saved but nothing else changed. The cost base is the same, the team is the same size and the savings exist mainly in the original slide.
That does not mean automation was a bad idea. It usually means the return on investment (ROI) was measured carelessly, both before and after.
Why automation ROI is so often overstated
Automation business cases tend to be optimistic in predictable ways. Knowing them makes it easier to build a case you can defend.
- No baseline. Nobody measured how long the task took, how often it happened or how many errors it produced before the change. Every later claim is a guess.
- Only the build cost is counted. Licences, maintenance, monitoring, changes and the time people spend handling exceptions are left out.
- Hours saved are treated as cash saved. Freeing thirty hours a month across five people does not reduce salaries. It creates capacity, which only has value if it is used.
- Volumes are assumed, not measured. The case assumes a thousand transactions a month when the real figure is closer to four hundred.
- Exceptions are ignored. The automation handles the standard case. The unusual cases, which may take most of the effort today, still need a person.
Step one: measure the baseline before you build
An honest ROI starts with an honest picture of today. For the process you plan to automate, measure for a few weeks:
- Volume: how many items go through the process each week or month, and how this varies across the year.
- Effort: how much hands-on time each item takes, including checking, correcting and chasing, not just the main task.
- Cycle time: how long an item takes from start to finish, including waiting.
- Errors: how often something goes wrong, and what it costs to fix, whether in staff time, refunds, penalties or lost customers.
- Exceptions: what share of items do not follow the standard path.
Simple methods are enough. A tally sheet for two weeks, a sample of twenty items timed from start to finish, or data already in your systems. The point is to have figures you can compare against later.
Step two: count every cost
The full cost of an automation over its life usually includes the following. Leaving any of these out makes the return look better than it is.
- Design and build: internal time and any external fees for analysis, development and testing.
- Licences and platform costs: workflow tools, connectors, integration services and any per-user or per-transaction fees, which tend to rise as usage grows.
- Running and maintenance: someone has to monitor the automation, fix it when a connected system changes and update it when the process changes. A modest annual allowance is realistic for most workflows.
- Exception handling: the time people still spend on items the automation cannot process.
- Change and training: the time needed for people to learn the new way of working.
Step three: separate the types of benefit
Benefits from automation come in different forms, and they should not be added together as though they were the same thing.
Cash benefits
Savings that appear in the accounts: overtime no longer paid, temporary staff no longer hired at peak, a planned hire avoided, penalties for late payments removed, early payment discounts captured. These are the most convincing and usually the smallest part of the case.
Capacity benefits
Hours released from routine work. They are real, but they become value only when they are redirected: to customer follow-up, to analysis, to absorbing growth without hiring. Write down in advance what the freed time will be used for, and check later whether it was.
Quality and speed benefits
Fewer errors, faster approvals, quicker responses to customers, better visibility for managers. These are often the most important benefits, and harder to turn into money. Measure them directly, as error rates and cycle times, rather than forcing them into a cash figure.
A worked example
Suppose a distribution company in Amman wants to automate the matching and approval of supplier invoices. Its baseline, measured over a month, shows about 600 invoices a month, with roughly six minutes of manual work each that the automation would remove. The figures below are purely illustrative. Amounts are in Jordanian dinars (JOD), and the hourly cost is “fully loaded”: salary plus social security and other employment costs.
| Item | Calculation | Amount per year |
|---|---|---|
| Hours released | 600 invoices × 6 minutes × 12 months | 720 hours |
| Value of hours, if redeployed | 720 hours × 6 JOD fully loaded hourly cost | 4,320 JOD |
| Licences and platform | 75 JOD per month | -900 JOD |
| Maintenance and support | Annual allowance | -500 JOD |
| Net annual benefit | 2,920 JOD | |
| One-off build cost | 5,000 JOD |
On these figures, the build cost is recovered in a little under two years, but only if the 720 hours are actually used for something valuable. If they are not, the cash return is negative, because licences and maintenance still have to be paid. The honest business case therefore rests on two things the table does not show: what the accounts team will do with the time, and the value of fewer payment errors and faster approvals, measured separately. Presented this way, the case is smaller but far more credible.
Step four: track results after go-live
Agree before go-live which measures you will check, and when. Three and six months after launch, compare volume, effort, cycle time, errors and exceptions with the baseline. Check whether licence costs match expectations and whether the freed time went where it was supposed to go.
Be prepared for mixed results. Some automations deliver more than expected, usually through quality improvements nobody predicted. Others deliver less, often because exceptions were underestimated. Both are useful lessons for choosing the next automation.
Returns you can stand behind
An honest ROI case is smaller than the one in a vendor proposal, but it survives a review by finance and tells you which automation to build next.
Our Intelligent Automation service starts every engagement with a measured baseline and ends with tracking against it. It reflects the final stage of the KEMLAT Method, Track, where impact is measured and followed rather than assumed.



