Automation opportunity assessment
How to find business processes worth automating
Asking a team what it wants automated rarely produces a useful project definition. People know the work, the delays and the workarounds, but they are not expected to know which intervention should replace them.
A better starting point is to examine a normal working week and listen for repeated steps, information copied between systems, and jobs people have already agreed need sorting out. Those signals create a shortlist. Evidence decides what happens next.
The search question
What business process should be automated first?
Look for several of these signals in the same process. One symptom may be an inconvenience, while a repeated pattern gives you something worth measuring.
- The same complaint appears in several roles or teams.
- Work waits in inboxes, spreadsheets or personal task lists.
- One person knows how to recover the process when it fails.
- Customers chase because nobody can see the current status.
- A proposed fix has been postponed because the process boundary is unclear.
Before and after
Technology-led question
- 1Where can we use AI?
- 2A tool is selected
- 3A task is found for it
Operational question
- 1What needs to change?
- 2The current work is measured
- 3Options are compared
This is a diagnostic sketch rather than a proposed system. The real route depends on the source data, decisions, exceptions and controls in the work.
A worked example
Example: proposal preparation takes three days
The first idea may be to automate proposal writing. Mapping the work might show that drafting takes 30 minutes while two days are lost waiting for survey notes, rate confirmation and senior approval.
A writing model would make the visible task faster without fixing the delay. The better first move may be a standard intake form, a current rate source and a clear approval threshold.
The calculation
30 minutes drafting + 2.5 days waiting is not a 30-minute problem
Active working time, elapsed time and rework need to be measured separately. Otherwise the project can make one step faster while the customer sees no improvement.
What to check
Define the work before choosing the technology.
Start with the person affected
Name who experiences the delay, duplication or error before describing the technology that might change it.
Measure the present position
Record volume, elapsed time, active time, rework and exceptions. Assumed savings are a poor baseline.
Test simpler interventions
Stop unnecessary work, remove hand-offs, standardise inputs and check fixed rules before adding AI.
Write down the stop condition
A useful test states what evidence would cause the team to change direction or end the project.
Reasons to stop or change direction
Not every visible problem needs automation.
- The proposed outcome has no owner.
- Nobody has measured the current performance.
- The problem is described only through a product or vendor name.
- The process boundary moves whenever the project is discussed.
- The business case counts all staff time as cash that will be removed.
Calculate your task
How much time is tied up in it?
Use one task and its normal weekly volume. Keep waiting time and correction work separate so they can be examined as well.
Based on 47 working weeks, 7.5 hours a day and 37.5 hours a week. This is capacity tied up in the task, not a promise of cash savings.
The next useful step
Write a one-page opportunity brief covering the affected people, unit of work, baseline, systems, consequence of error, required oversight, success measure and stop condition. It may not need the words AI or automation anywhere on the page.
Readiness assessment
Answer seven questions to see whether the business is ready to test, build or first define the operational problem more carefully.
Common questions
Straight answers.
How do I identify processes to automate?
Examine a normal working week for repeated steps, information copied between systems, recurring delays, manual reconciliation and work that depends on one person. Measure the current process before deciding how it should change.Which business process should be automated first?
Start with a process that has a defined owner, measurable volume, stable inputs, a clear outcome and manageable consequences when something goes wrong. High visibility alone does not make a process a good first candidate.How do I calculate the value of automation?
Measure active working time, elapsed time, rework, error cost and demand volume. Time released is capacity rather than guaranteed cash, so the business case should state how that capacity will be used.
Bring the awkward examples
Talk through one process before choosing a product.
A few recent cases, the current systems and the person who handles the exceptions are enough for a useful first conversation.
Book a call