top of page

How Process Mining Shows You What to Automate First

8 hours ago
7 min read

Ask five managers how invoice approval works and you will often get five slightly different answers, plus a tidy flowchart from a workshop two years ago. Then look at what the systems actually recorded. The same process may run dozens of different ways, with invoices looping back for corrections, approvals skipped under pressure and files sitting in one team's queue for days.


Automating from the flowchart would mean automating a process that does not really exist. Process mining shows you the real one, and with it, where automation will pay off first.


Quick answer: Process mining analyzes the event logs your systems already record to show how a process really runs: every path, delay, loop and deviation. To decide what to automate first, it highlights high-volume, repetitive steps with heavy manual effort and few exceptions, and separates them from steps that need fixing or redesign before automation. CMEx uses process mining as the first step in its Intelligent Automation practice.


What is process mining?


Gartner defines process mining as a technique designed to discover, monitor and improve real processes, not assumed ones, by extracting knowledge from the event logs of information systems.


An event log is simply a record of what happened, to which case, and when. Every ERP, CRM, core banking, ticketing or workflow system keeps one. A simple example from an invoice process looks like this:


Case ID

Activity

Timestamp

Performed by

INV-1042

Invoice received

3 Mar, 09:14

Email intake

INV-1042

Invoice data entered

4 Mar, 16:40

AP clerk

INV-1042

Price mismatch flagged

4 Mar, 16:41

ERP

INV-1042

Invoice corrected

7 Mar, 11:05

AP clerk

INV-1042

Invoice approved

9 Mar, 15:22

Finance manager

Process mining takes thousands or millions of rows like these and rebuilds the real process from them.


What does process mining reveal?


Process mining usually works in three ways:


  • Discovery: it builds a map of how the process actually runs, including every variant, without relying on anyone's assumptions.

  • Conformance checking: it compares the real process with the intended one, showing skipped approvals, wrong sequences and other deviations that matter for audit and compliance.

  • Enhancement: it adds measures such as waiting time, rework and cost to the map, so you can see exactly where time and money are lost.


The results often surprise teams. In research published in Communications of the ACM, Wil van der Aalst, a founder of the field, notes that applying process mining in hundreds of organizations shows that managers and users tend to overestimate their knowledge of their own processes.


Why start with process mining before automating?


  • You avoid automating a broken process. Automation makes a process faster, including its waste. Our guide on RPA vs BPM vs AI explains why the order of decisions matters more than the tool.

  • You get a real baseline. Cycle time, rework rate and cost per case are measured from data, so you can prove the improvement later.

  • You build a stronger business case. Decisions rest on evidence from the systems, not on opinions from workshops.

  • You see compliance gaps early. Deviations that would be automated and repeated thousands of times are found and fixed first.


How does process mining show what to automate first?


Process mining turns each part of a process into measurable signals. These signals show which steps are ready for automation and which need work first.

Signal from the data

What it tells you

What it means for automation

High volume

The step runs thousands of times

Even small time savings add up quickly

One dominant path

Most cases follow the same sequence

Rules are clear, so automation is reliable

Many manual touches

People retype, copy or check data by hand

A strong candidate for RPA or integration

Long waits between teams

Work sits in queues between handoffs

Workflow and routing (BPM) will help more than bots

Frequent rework loops

Cases go back for corrections

Fix the root cause before automating

High exception rate

Many cases do not fit the standard path

Simplify first, or keep people in the loop

Unstructured inputs

Data arrives in emails, PDFs or scans

Document intelligence is needed before other automation

Compliance deviations

Required steps are skipped or reordered

Redesign the controls, then automate them

A simple way to rank candidates: automate first where volume is high, one path dominates and manual effort is heavy. Fix first where rework and exceptions are high. Redesign the flow where most of the time is spent waiting rather than working.


What does this look like in a real process?


Take a purchase-to-pay process. Process mining might show that:


  1. Most invoices follow one standard path, but a large share loop back because of price mismatches between the invoice and the purchase order.

  2. AP clerks rekey invoice data from PDFs into the ERP for most suppliers.

  3. Invoices wait several days for approval because they sit in managers' inboxes.

  4. A small number of high-value invoices skip the second approval required by policy.


That points to a clear order of work. Fix the price mismatch at its source by improving purchase order data. Use document intelligence to read invoices and remove rekeying. Add workflow routing and reminders to cut approval waits. Enforce the second approval in the workflow so it cannot be skipped. Only then add bots for any remaining repetitive steps.


What is the difference between process mining and task mining?


Process mining looks at the whole process across systems, using event logs. Task mining looks at what individual users do on their screens, such as clicks, copying and switching between applications, within a single step. Process mining shows where to look. Task mining then shows exactly how a manual task is done, which helps when designing an RPA bot for it.


What data does process mining need?


At minimum, each event needs three things: a case ID that links events to the same case, an activity name and a timestamp. A resource or user field adds useful detail. Common sources include ERP, CRM, core banking, claims, ticketing and workflow systems.


The hardest part is usually preparing the data: finding the right tables, linking cases across systems and cleaning timestamps. Planning for this early keeps the project on schedule.


Where is process mining most useful?


  • Banking & Financial Services: loan origination, KYC and account opening, where waiting time, rework and compliance deviations are costly.

  • Insurance: claims handling, where variants and handoffs drive cycle time and customer satisfaction.

  • Public Infrastructure and the public sector: high-volume applications and approvals that must follow strict rules.

  • Retail: order-to-cash and returns, where exceptions and manual checks slow refunds and fulfillment.

  • Energy & Utilities: maintenance work orders, where delays between planning, approval and execution affect uptime.


How do you get started with process mining?


  1. Choose one process with a clear business pain, such as slow approvals or high rework.

  2. Identify the systems that record its events and the owners who can give access.

  3. Extract and prepare the event log, with case IDs, activities and timestamps.

  4. Discover the real process and measure volume, waiting time, rework and deviations.

  5. Rank the opportunities using the signals above: automate, fix or redesign.

  6. Act, then measure again against the baseline, and keep monitoring the process over time.


How CMEx approaches process mining and automation


CMEx's Intelligent Automation practice starts from one rule: re-engineer the underlying process, instrument it, and only then automate. Process mining is how that first step is grounded in evidence.


In practice, that covers:


  • Process mining to map the real process and rank automation opportunities.

  • Business workflow and decision automation to fix handoffs, approvals and routing.

  • Document intelligence and content services for processes that run on emails, forms and scanned documents.

  • Robotic process automation through CMEx's strategic alliance with AutomationEdge, used only where it fits.

  • An Intelligent Automation Center of Excellence focused on process mining, RPA, content services and decision automation, engineered for audit.


Frequently asked questions


How is process mining different from process mapping workshops? 


Workshops capture how people believe a process works. Process mining uses system data to show how it actually works, including every variant, delay and deviation. The two work well together: data shows what happens, and people explain why.


Do we need special software for process mining? 


Yes. Process mining platforms connect to your systems, build the event log and create the process maps and measurements. Several enterprise platforms are available, and the right one depends on your systems, data volumes and existing automation tools.


How long does a first process mining project take? 


A focused project on one process typically takes weeks rather than months. Most of the effort goes into finding, extracting and cleaning the event data. Once that pipeline exists, monitoring can continue with far less effort.


Is process mining only useful for automation? 


No. It also supports compliance and audit, cost reduction, customer experience and process redesign. Many organizations keep it running after automation to confirm the gains hold and to spot new problems early.


Does process mining raise data privacy concerns? 


It can, because event logs may include personal or employee data. Good practice is to collect only the fields needed, mask or pseudonymize personal data, control access to results and focus analysis on processes rather than individual performance.


What is the best process to start with? 


Choose a high-volume process with a visible pain point and good system data, such as invoice processing, loan or claims handling, or order-to-cash. A clear problem makes the value easy to show.


Can process mining measure the results of automation? 


Yes. Because it uses the same event data before and after a change, it can show exactly how cycle time, rework, waiting time and compliance have shifted once automation is live.


Conclusion


The biggest automation mistakes start with the wrong process. Process mining replaces assumptions with evidence: it shows how work really flows, where time is lost and which steps are truly ready for automation. Start there, fix what is broken, and automate the rest with confidence.


Not sure which processes to automate first? A CMEx practice principal, not a sales rep, will review one of your processes and show where the evidence points.



 
 
bottom of page