How to Build a Business Case for Process Automation

A business case for process automation gets approved when it answers three questions in order: what the problem costs today, what will change, and how you will prove it changed. Most cases that fail skip the first question and open with the technology. This guide sets out the order that works, the numbers a finance director will test and the mistakes that send a case back.

Start with one process, not a programme

A case for “automation across operations” asks the board to fund a direction. A case for one process asks them to fund a result. Pick the process where the pain is already visible to someone senior: orders that leave late, invoices that wait for approval, claims that are handled twice.

Three tests help you choose:

• It runs often enough to measure. A few hundred cases a month at least.
• It leaves a record. Its steps pass through a system that logs when each one happened.
• It has an owner. A process with no owner has nobody to sign off the change.

Price the cost of doing nothing

This is the section finance reads first, so build it from figures they can check. For the process you chose, count four things over the last three months:

• Volume. How many cases ran each month.
• Delay. How long each case waited between steps, and what a day of waiting costs in penalties, cash tied up or lost orders.
• Rework. How many cases were handled more than once, and the staff time each repeat took.
• Failure. How many cases missed their deadline or were written off.

Multiply each count by its unit cost and add them up. Use your own figures; the arithmetic matters more than the example below.

Where you cannot count something, say so. A case that marks one number as an estimate is trusted more than one that presents every number as certain.

Set a target you can measure in 90 days

A target of “greater efficiency” cannot be missed, which is why nobody believes it. Name one measure, its value today and its value after the change. “Cut the time from order to dispatch from five days to four” is a target. It can be checked on a date, by someone other than you.

Keep the first target modest. A cautious target you beat funds the second project, while an ambitious one you miss can end the programme.

The three numbers finance will test

Show how you reached each figure. The working is set out in our guide to payback times and cost drivers, and you can test your own numbers in the ROI calculator.

Count the whole cost

The licence or the consultancy fee is the cost everyone includes. Three others are usually left out, and finance will find them:

• The work of extracting and cleaning the data, which often takes the first few weeks.
• The time your own people give to workshops, testing and acting on what is found.
• The cost of running the new process once the project team has gone.

Put all of them in. A case that is honest about its costs and still pays back in under a year is far stronger than one that looks cheaper and is later found to have missed a line. For what each of those lines typically involves, and the licence prices vendors publish, see what a process mining implementation costs.

Name the risks before someone else does

Every reviewer asks what could go wrong. Answer it in the case itself, with what you will do about each risk. The common ones are poor data quality, staff working around the new process and a result that fades once attention moves on.

The last of these deserves a line of its own. Say who will own the measure after the project, and how often it will be checked. Monitoring the process on live data is what stops an improvement being quietly reversed.

Decide how the work gets done

The case should say whether you will build the capability in-house, buy a platform or bring in a specialist firm, and why. The three routes differ most in how long they take to a first useful result and in who carries the running cost. The trade-offs are compared in build vs buy vs hire.

What the finished case looks like

Two pages is enough, plus an appendix:

• Page one. The process, what it costs today, the target and the date it will be measured.
• Page two. The costs, the three numbers above and the risks.
• Appendix. The working, for anyone who wants to check it.

If you cannot fill the first page, you are not ready to ask for money. That is useful to know before the meeting, and it usually means the next step is a short piece of discovery work to get the numbers.

Related: see what should a process improvement consultant deliver in the first 90.

Frequently asked questions

Related reading

• Process mining ROI: payback times and cost drivers
• ROI calculator
• Build vs buy vs hire