Mapping the invisible tax: How to spot workflows worth automating

The biggest operational bottlenecks aren't broken systems—they're the manual workarounds your team has accepted as normal.

Every business has a version of this story.

A customer sends an email. Someone copies the details into a spreadsheet. Another person updates the CRM. A third sends a message to confirm the request has been received. Later, someone checks whether the first person actually completed their part.

Nothing is technically broken. The email arrived. The spreadsheet works. The CRM is running. Everyone has access to the tools they need.

And yet, a simple request has passed through four people, three systems and a handful of small tasks before the business has meaningfully moved forward.

Ask the team how work gets done, and they might tell you everything is fine.

They have a process for it.

That phrase deserves a closer look.

Sometimes, we have a process for it means a task is well understood and handled efficiently. Other times, it means somebody has spent the last two years doing something tedious every morning because nobody has questioned whether it needs to happen that way.

The employee has adapted. The team has adapted. The business has adapted. And somewhere along the way, an avoidable problem became part of the job.

1. The workaround is often the first clue

When leaders look for operational inefficiencies, they tend to look for obvious failures: a system that keeps crashing, an approval that takes three weeks, a department that cannot meet its targets.

Those problems matter. But they are not the only places where time and money disappear. Some of the most expensive inefficiencies are hidden inside processes that appear to work perfectly well.

Consider a finance team that downloads transactions from a banking portal, cleans up the spreadsheet, matches entries against invoices and then manually updates an accounting system. The work gets done every month. The reports are produced. Nobody raises an alarm.

But the process depends on people moving information between systems, checking for mistakes and correcting inconsistencies that a better-connected workflow could prevent.

Or consider a diagnostic centre where patient details arrive through different channels, staff re-enter information into internal systems, radiologists are notified separately, and referring doctors must follow up to obtain reports. The team might be hardworking and competent. The underlying process can still create unnecessary delays and opportunities for error.

These are not necessarily failures of people or even failures of individual software products. They are failures of the connections between tasks, teams and systems.

When a process depends on someone remembering to move information from one place to another, the person may not be the problem. The process is asking them to do work the system could handle.

THE INVISIBLE TAX ACCUMULATION CHAIN:

Customer Request Ingress (Email / Portal)
         |
    [ Manual Step 1 ] ----> Operator copies customer details into Intake Tracker
         |
    [ Manual Step 2 ] ----> Operator cross-checks inventory in ERP
         |
    [ Manual Step 3 ] ----> Status update typed into CRM
         |
    [ Manual Step 4 ] ----> Slack message sent to Fulfillment Team
         |
    [ The Latency Tax ] --> Customer waits 48 hours for a 2-minute decision.
                               Data discrepancies emerge in 1 out of 5 orders.

2. The cost isn't just the minutes on the clock

Suppose an operations employee spends 12 minutes each morning transferring information between two systems. It doesn't sound like much.

Across 22 working days, that's 4.4 hours a month for one person.

Now imagine eight people doing similar work, each spending those same 12 minutes a day. Collectively, the business is spending about 35 hours a month on that one category of manual work. That's almost a full working week for one employee, every month, spent on a task that could be automated or eliminated.

And we haven't counted the time spent correcting mistakes, chasing missing information, answering questions about the process, or waiting for the next person to pick things up.

Research from McKinsey has repeatedly highlighted an essential distinction: automation potential is best assessed at the level of individual work activities, not entire job titles. A role contains a mixture of routine tasks, specialist judgement, communication, and strategic decisions.

Don't ask, "Which department should we automate?"

Ask, "What are people in this department repeatedly doing that doesn't require their full attention, expertise, or judgement?"

The answers tend to be far more revealing.

3. How to find the work nobody thinks to report

You will rarely find the best opportunities by sending a company-wide email asking employees which processes should be automated. You may get a few surface complaints, but you'll miss the work people have accepted as normal.

Instead, look for the four structural signals:

A. Follow the information

Pick a common business event: a new customer, an incoming order, a supplier invoice, a patient referral, or a support request. Follow it from the moment it enters the business until the work is complete.

• How many times is the same information entered or copied?
• Which systems or people must receive it before the next step can begin?
• Who checks whether somebody else has completed their part?
• Which steps exist only because two systems cannot communicate?

Automating the copy-and-paste step without fixing the underlying connection only makes an awkward process slightly faster.

B. Find the spreadsheets that have quietly become essential

Spreadsheets are flexible and useful. The warning sign is not that a spreadsheet exists; it's that the business depends on one that sits outside its formal systems and has become essential to daily operations.

• What would stop working if the person maintaining this file were absent for two weeks?
• Does anyone independently verify the information it contains?
• Why does the file need to exist alongside the systems the business already pays for?

The objective is not to eliminate spreadsheets. It is to understand why the business depends on them.

C. Listen for the phrases people use to explain delays

Some of the most useful clues come from ordinary workplace conversations:

• "I'm waiting for them to send it."
• "I'll just update both systems."
• "Let me check the other spreadsheet."
• "We usually have to call to confirm."
• "That's how we've always done it."
• "Don't worry, I have my own way of tracking those."

Each phrase points towards an undocumented dependency, duplicate entry, or information silo.

D. Look for work that waits on people rather than decisions

Not every delay is waste. Some work should wait for approval, professional judgement, or strategic deliberation. But there is a meaningful difference between waiting for someone to make a decision and waiting for someone to notice that a decision is needed.

An invoice might sit untouched because the manager has not been notified. A customer request might languish in a shared inbox because nobody has been assigned responsibility. Good automation removes unnecessary handling without removing necessary human judgement.

4. Not every repetitive task deserves automation

Once you've identified candidates, it's tempting to start with whichever task looks easiest to automate. Resist that urge. A task can be repetitive and still be a poor investment. It may happen too rarely, depend on poor source data, or involve ambiguous decisions where mistakes carry severe risks.

Evaluate candidate workflows across eight dimensions:

1. 1
   Frequency: How often does the task happen?
2. 2
   Time: How much staff time does it consume in total?
3. 3
   Error rate: How often is information entered incorrectly, missed, or duplicated?
4. 4
   Delay: Does the task hold up a customer, colleague, or downstream workflow?
5. 5
   Business impact: What does the delay or mistake cost the enterprise?
6. 6
   Complexity: Does the task follow predictable rules, or require deep human discretion?
7. 7
   Risk: What happens if the system makes an unhandled error?
8. 8
   Feasibility: Can existing APIs support the change, or is a purpose-built system required?

Assessing candidate workflows:

• Copying order details into a tracker (Strong Candidate): Frequent, rule-based, and verifiable against source data.
• Routing incoming support requests (Worth Testing): Can be classified and routed automatically with confidence scoring, using fallback queues for ambiguous requests.
• Approving high-value payments (Keep Oversight): Information gathering and compliance verification should be automated, but final authorization authority requires human controls.

5. The opportunities hiding in plain sight

Once you evaluate activities rather than departments, five high-leverage automation patterns emerge:

1. 1
   Moving information between systems: Customer records, orders, invoices, appointments, and status updates that travel manually between tools.
2. 2
   Chasing updates and sending reminders: Automated deadline alerts, status syncs, and proactive escalation rules.
3. 3
   Checking documents and entering records: Extracting fields, validating required attachments, and preparing records for review.
4. 4
   Routing requests and coordinating handoffs: Real-time assignment engines that dispatch work to the right specialist based on workload and skill.
5. 5
   Compiling reports and checking performance: Automated ETL pipelines that assemble unified operational analytics from fragmented silos.

Not every workflow needs an AI agent. Sometimes a simple integration or a well-designed state machine is cheaper, easier to maintain, and more dependable. Solve the problem with the least unnecessary complexity.

6. Before you automate, understand why the work exists

There is a trap in operational improvement: seeing a manual step and immediately deciding to eliminate it.

Sometimes that step is compensating for a broken process. Sometimes it exists because the software doesn't fit the way the business operates. And sometimes it is there for an essential regulatory reason that was never documented.

Automating a broken process only makes errors move through your business at digital speed.

Before building anything, map the workflow as it actually happens—not as the employee handbook claims it happens.

Identify who starts the work, what information they need, which systems are involved, where decisions are made, what happens when an exception occurs, and who is accountable for the final outcome. Then ask what the process should look like if you designed it from scratch today.

Start with one week of observation

You don't need to begin with a multi-million-dollar technology project. Start by making the invisible visible:

• Monday: Ask each team to identify the recurring tasks that frustrate them most.
• Tuesday: Follow two or three core workflows end-to-end. Record every handoff, duplicate entry, approval, and workaround.
• Wednesday: Estimate the time involved, frequency, and downstream cost of errors or delays.
• Thursday: Rank the opportunities by business value, feasibility, and risk. Choose one contained process to test.
• Friday: Define what success looks like—including the baseline metrics, safeguards, and operational owners.

The spreadsheet nobody can retire. The customer request that needs three follow-up messages. The employee who knows which system to update first because the official tool never works. These are not merely inconveniences—they are the roadmap to your next competitive advantage. Your team shouldn't have to spend its time compensating for the limitations of its software.

At ZeroComplex, we help organizations eliminate the invisible tax by architecting software and autonomous workflows around the work that actually needs to happen. Whether that means replacing brittle manual data pipelines with unified domain models, orchestrating real-time exceptions across legacy tools, or engineering a dedicated operational system from scratch, we build software that removes administrative drag and empowers teams to execute at full speed.