"We've already looked at our processes. There's nothing major to automate." — Brisbane engineering firm, before we found $127,000 in annual waste
When we ran the audit, we found 11 processes costing them over $10,000 each per year. The owner had estimated maybe two or three.
Every business has hidden waste. Tasks that take 10 minutes but happen 40 times a day. Processes that require 3 people when 1 would do. Data that gets copied from one system to another by hand.
Most businesses never find it because they don't know how to look. Here's the exact framework I use to identify automation opportunities—the same one that uncovered $127,000 at that Brisbane firm.
The 5-Question Process Audit
Every process audit starts with the same five questions. These aren't complicated, but most businesses have never asked them systematically.
Question 1: How often does this happen?
Daily? Weekly? Monthly? The frequency determines whether automation is worth the effort. A task that happens twice a month isn't worth automating. A task that happens 50 times a day is costing you a fortune.
Question 2: How long does it take each time?
Time the actual work, not the calendar time. An invoice might sit in a queue for two days, but the actual data entry takes 6 minutes. We care about the 6 minutes.
Question 3: Who does it?
This determines the labor cost. A $25/hour admin copying data is different from a $75/hour specialist doing the same thing. Both are worth examining, but the specialist's time is more expensive to waste.
Question 4: What triggers it?
Does something specific start this process? An email arrives, a form gets submitted, a date passes, a threshold gets hit. If you can define the trigger, you can automate the response. If the trigger is "someone remembers to do it," that's a different problem.
Question 5: What can go wrong?
Every process has exceptions. Invoices with missing data. Forms filled out incorrectly. Edge cases that need human judgment. Understanding these upfront tells you how much of the process can be automated versus how much needs human oversight.
Run every process through these five questions. Write down the answers. You'll start seeing patterns immediately.
The Automation Candidate Scorecard
Not every process is worth automating. Some are perfect candidates. Others need to stay manual, at least for now.
Here's how to score each process.
Volume and Frequency
How often does this happen? Daily or hourly tasks score high. Monthly or occasional tasks score low. You need enough volume to justify the setup.
Repetitiveness
Does it happen the same way every time, or is every instance different? The more consistent the process, the easier it is to automate. If every case is unique, you're looking at AI assistance rather than full automation.
Rule-Based Logic
Can you describe the decisions as if/then statements? If the invoice is under $500, auto-approve. If over $500, route to manager. Clear rules mean clear automation. Fuzzy judgment means human-in-the-loop.
Process Stability
Has this process been stable for at least 6 months, or does it change constantly? Stable processes are easier to automate. Constantly changing processes need flexible AI, not rigid workflows.
Digital Inputs
Is the data already digital and structured, or are you dealing with paper, handwriting, or unstructured emails? Digital and structured is easier. Paper and chaos require more preprocessing.
Error Rate
How often do mistakes happen in this process? High error rates actually make processes better candidates for automation. If humans are making mistakes, machines will be more consistent.
Score each criterion from 1 to 5. Add them up.
25 or higher: Strong automation candidate. Start here.
15 to 24: Partial automation possible. Automate what you can, keep humans for the rest.
Below 15: Keep manual for now. Revisit when the process matures.
The ROI Calculation
This is the math that gets projects approved. It's simple, but most people don't do it.
The Formula
Take the time per task in minutes. Divide by 60 to get hours. Multiply by how many times it happens per day. Multiply by 250 working days per year. Multiply by the fully loaded hourly rate.
That's your annual cost for one manual process.
Fully loaded hourly rate means salary plus benefits plus overhead. A good rule of thumb is base hourly rate plus 35%. Someone making $25/hour actually costs you around $34/hour when you factor in everything.
A Real Example
Invoice data entry. 6 minutes per invoice. 40 invoices per day. $35/hour fully loaded rate.
6 divided by 60 equals 0.1 hours per invoice. 0.1 times 40 invoices equals 4 hours per day. 4 hours times 250 days equals 1,000 hours per year. 1,000 hours times $35 equals $35,000 per year.
That's $35,000 annually on one task. One person. One process.
Now multiply that across your organization. How many people do similar tasks? How many processes have you never measured?
The compound math is what gets executives to pay attention. 6 minutes doesn't sound like much. $35,000 does. And most businesses have dozens of these hidden costs.
Common Edge Cases and How to Handle Them
When you start auditing processes, you'll hear the same objections repeatedly. Here's how to think about each one.
"We've always done it this way"
This isn't a blocker. It's an opportunity. Most processes have never been questioned. When you audit them, you often find steps that made sense five years ago but don't anymore. Workarounds that became permanent. Entire tasks that could be eliminated, not just automated.
Before automating, ask whether this process should exist at all. Sometimes the answer is to remove it entirely.
"We can't explain all the exceptions"
People handle exceptions when they hit them. They just can't list them all upfront. That's normal.
Build guardrails. When the automation hits something unexpected, route it back to a human. Log what happened. Over time, the system learns the exceptions and handles more of them automatically. You don't need to know every edge case on day one.
"The process changes all the time"
This is only a problem with old-school automation that hard-codes linear steps. Modern AI adapts.
Use AI-based automation, not rigid if/then workflows. Design for flexibility from the start. The system should handle variations, not break on them. If your automation can't handle change, you've built the wrong kind of automation.
"It requires human judgment"
Most processes are 80% repetitive and 20% judgment. Automate the 80%.
Human-in-the-loop means AI handles the grunt work, surfaces the decisions that need a human, and waits for approval before proceeding. Humans stay in control of what matters. They just don't waste time on what doesn't.
"The team doesn't want to talk about it"
They're worried about their jobs. That's a legitimate fear, and ignoring it kills projects.
Address it directly. Automation usually means more interesting work, not redundancy. Show them what their role becomes. Less data entry, more judgment calls. Get them involved in designing it. People don't resist change they helped create.
The Real Red Flags
The edge cases above are solvable. These next two are the actual project killers, and they're both about people, not technology.
Knowledge as Job Security
When employees are evasive about how they do their work, it's rarely malicious. The process is their value. It's what makes them irreplaceable. Sharing it feels like giving away their leverage.
Signs to watch for: Vague answers like "it depends" or "I just know." Resistance to documenting steps. Claims that it's too complicated to explain. Being the only person who can do it and liking it that way.
This doesn't mean walk away. It means the conversation isn't about automation yet. It's about trust and job security. Address that first. Show them that capturing their knowledge makes them more valuable, not less. They become the expert who designs the system, not the person who gets replaced by it.
No Management Appetite for Change
Automation projects don't fail because of technology. They fail because someone has to enforce new ways of working, and nobody wants that job.
Signs to watch for: Leadership says "sounds great" but won't allocate time or resources. No one owns the outcome. "Let's just try it and see" with no real commitment. Previous automation projects that stalled after the demo.
If management won't sponsor the change, the project is dead before it starts. The technology works. The humans won't use it. Without someone with authority driving adoption, you're building software that collects dust.
The Prioritization Matrix
Once you've audited your processes and scored them, you need to decide what to tackle first.
Plot each candidate on two axes: effort to implement and impact on the business.
Low effort, high impact: Do these first. These are your quick wins that build momentum and prove the concept.
High effort, high impact: Plan these carefully. They're worth doing but need proper resourcing.
Low effort, low impact: Fill time with these. Nice to have but not priorities.
High effort, low impact: Skip these entirely. Not worth the investment.
Start with two or three quick wins. Get them working. Prove the ROI. Then use that success to fund the bigger projects.
The businesses that succeed with automation don't start with their most complex process. They start with something achievable, deliver results, and build from there.
Bottom Line
You now have the same framework I use with clients. The five questions. The scorecard. The math. The edge cases. The red flags.
Pick one process this week. Run it through the five questions. Score it. Calculate the cost.
Most people are surprised by what they find. The waste is there. It's just never been measured.
That Brisbane engineering firm? They started with their highest-scoring process—purchase order approvals. Automated it in 3 weeks. Saved $31,000 in the first year. Then used that win to fund the next four projects.
Want help running an automation audit? We'll identify your highest-value automation opportunities and show you exactly where the waste is hiding. Start your process analysis →