RPA Process Discovery
Process Discovery is the activity of finding, mapping, and evaluating business processes to identify the best candidates for RPA automation. Before you build a single bot, you must understand exactly what tasks employees perform, how long those tasks take, and which ones follow consistent rules.
Think of Process Discovery as detective work — you gather clues from people, systems, and data to build a complete picture of how work actually happens.
Why Process Discovery Is Critical
Many RPA projects fail not because of bad bots but because of poor process understanding. If you automate a broken or undocumented process, the bot inherits all the problems of that process. Discovery ensures you automate the right things in the right way.
Three Methods of Process Discovery
Method 1: Interviews and Workshops
You meet with department managers and frontline employees to understand what tasks they perform. You ask questions like:
- What tasks do you repeat every day?
- Which tasks take the most time?
- Which tasks have the most errors?
- What data do you work with — spreadsheets, emails, web portals?
- Are there steps that you feel should not require a human?
Method 2: Process Observation
You sit with an employee and watch them perform the task in real time. You note every click, every application they use, every decision they make, and every exception they handle. This reveals steps that employees do automatically without thinking — steps they may forget to mention in an interview.
Method 3: Task Mining (Automated Discovery)
Task Mining software runs on employees' computers and records their interactions — which applications they use, which screens they visit, and how long each step takes. The software analyses this data and generates a process map automatically. UiPath Task Mining and Automation Anywhere's Discovery Bot are examples of this approach.
Comparison: Three Discovery Methods
| Method | Time Required | Accuracy | Best For |
|---|---|---|---|
| Interviews | Low | Moderate (relies on memory) | Initial scoping |
| Observation | Moderate | High (you see real steps) | Detailed mapping |
| Task Mining | High setup, then automated | Very high (data-driven) | Large-scale discovery |
The Process Discovery Workflow
STEP 1: IDENTIFY SCOPE
Decide which departments or functions to investigate
Meet with business leaders to prioritise areas
│
▼
STEP 2: COLLECT INFORMATION
Conduct interviews with process owners
Observe employees performing tasks
Collect existing documentation
│
▼
STEP 3: MAP THE PROCESS
Draw a step-by-step process map
Include all applications, decisions, and data sources
Note where exceptions occur
│
▼
STEP 4: MEASURE THE PROCESS
Calculate: How many times per day/week/month?
Calculate: How long does each step take?
Calculate: What is the current error rate?
│
▼
STEP 5: EVALUATE AUTOMATION POTENTIAL
Score each process on the RPA suitability criteria
Estimate time savings and ROI
│
▼
STEP 6: PRIORITISE AND PRESENT
Rank candidates by impact vs complexity
Present shortlist to stakeholders for approval
Process Mapping Techniques
Swimlane Diagram
A swimlane diagram shows which person or system is responsible for each step in a process. It makes handoffs between people and systems visible. This is particularly useful in RPA because it highlights which steps a bot can take over and which steps still need a human.
INVOICE APPROVAL SWIMLANE
──────────────────────────────────────────────────────
VENDOR: [Submit Invoice] ──────────────────────────
│
ACCOUNTS ─────────────────────── │ ─────────────────
PAYABLE: [Receive Invoice]◀───── │
│ │
▼ │
[Enter in ERP] │
│ │
MANAGER: ──────│─────────────────│──────────────────
│ │
[Review & Approve] │
│ │
ACCOUNTS ──────│─────────────────────────────────────
PAYABLE: ▼
[Process Payment]
│
▼
[Notify Vendor]
─────────────────────────────────────────────────────
BOT SCOPE: Enter in ERP + Notify Vendor (rule-based steps)
Key Metrics to Capture During Discovery
- Transaction Volume: How many times is the process performed per day, week, or month?
- Average Handle Time (AHT): How long does one instance of the process take?
- Full-Time Equivalent (FTE): How many employees work on this process and for how many hours?
- Error Rate: What percentage of transactions have errors that require rework?
- Seasonal Peaks: Are there times when volume spikes (month-end, year-end)?
FTE Calculation Example
Process: Invoice data entry Volume: 200 invoices per day AHT: 6 minutes per invoice Total daily effort: 200 × 6 = 1,200 minutes = 20 hours FTE needed: 20 hours ÷ 8 hours/day = 2.5 FTE After RPA: Bot AHT: 1 minute per invoice Total daily bot effort: 200 minutes = 3.3 hours FTE equivalent: 0.4 FTE FTE saved: 2.5 – 0.4 = 2.1 FTE
Process Discovery Output: The Automation Pipeline
The output of process discovery is an Automation Pipeline — a prioritised list of automation candidates, each with a brief description, estimated ROI, complexity rating, and recommended approach.
| Process | Dept. | Volume/Day | FTE Saved | Complexity | Priority |
|---|---|---|---|---|---|
| Invoice Entry | Finance | 200 | 2.1 | Low | 1 |
| Employee Onboarding | HR | 15 | 0.8 | Medium | 2 |
| Bank Reconciliation | Finance | 1 | 1.2 | Medium | 3 |
| Customer Account Update | CS | 300 | 1.5 | Low | 4 |
Summary
Process Discovery identifies, maps, and evaluates business processes before automation begins. It uses interviews, observation, and task mining software to understand how work happens. Key outputs include process maps, metrics (volume, AHT, FTE), and a prioritised automation pipeline. Thorough discovery prevents the costly mistake of automating the wrong processes or building bots on top of broken workflows.