MuleSoft RPA & Intelligent Document Processing (IDP) (Course #rpa-idp)

Programme every line is something you do

5 modules · 15 topics · in course-flow order

1. Module 1

   Where no API exists

   Objective. Diagnose which work belongs to MuleSoft RPA, to MuleSoft IDP or to an ordinary integration, and produce a scored candidate list.

   • Place RPA Manager, RPA Builder, RPA Recorder and RPA Bot on one topology diagram, marking the cloud control plane against the 3 on-premise components
   • Score 6 candidate processes on volume, rule stability and screen churn, keeping only those above 500 runs/month
   • Refuse to automate what an API already serves: audit the estate first and cap RPA scope at the 20% of steps no connector can reach
2. Module 2

   Record, then rebuild

   Objective. Construct a tested RPA Builder process from a recorded session and prove it survives a screen change.

   • Capture a session with RPA Recorder, then rebuild it as explicit action steps in RPA Builder — target: zero raw coordinate clicks
   • Split a 60-step process into named sub-processes with typed variables so a failed step resumes without replaying the whole run
   • Handle failure on purpose: per-action-step timeouts, retries capped at 3, and detailed trace logging switched on for the failing branch only (RPA Builder 6.7)
3. Module 3

   Attended, unattended and queued

   Objective. Design the bot fleet and run-configuration policy that meets the throughput target without colliding sessions.

   • Choose console/RDP sessions for attended work against secure sessions for unattended runs: 1 process at a time versus many in parallel
   • Deploy scheduled, invocable and on-demand run configurations, setting queue priority on the 1–200 scale so month-end pre-empts the overnight batch
   • Hold application logins as encrypted RPA Manager credentials rotated every 90 days, never inside a process step
4. Module 4

   Structured JSON from scanned paper

   Objective. Construct a published IDP document action that returns typed JSON and routes weak extractions to a human reviewer.

   • Build a document action on the pre-built invoice schema, then add up to 30 prompts for the fields the schema does not carry
   • Set the per-field confidence threshold at 90% and mark required fields, so a miss queues the document for review instead of posting bad data
   • Size intake against the real ceilings — 15 MB and 150 pages per API request, 300 requests/minute — and plan around the 24-hour result retention window
5. Module 5

   Hand the result back to APIs

   Objective. Prove the full chain — IDP extraction, RPA keystrokes, Mule flow — against the latency and human-touch budget.

   • Publish the document action to Anypoint Exchange and call the generated IDP API from a Mule flow, polling results on the documented 10-second interval
   • Invoke an RPA process with an API key against the RPA API URL (/rpa/api/v1), sync under a minute and async with polling beyond it
   • Prove the round trip end to end: scanned PDF to posted ERP reference in under 10 minutes, with under 15% of documents touched by a human

References & Official Documentation every claim above has a source

Module 1

• Considering Which Processes to AutomateMuleSoft Docs
• RPA OverviewMuleSoft Docs

Module 2

• RPA Recorder OverviewMuleSoft Docs

Module 3

• Deploying AutomationsMuleSoft Docs

Module 4

• Analyzing Documents Using Pre-Built SchemasMuleSoft Docs
• IDP Quotas and LimitsMuleSoft Docs

Module 5

• Invoking RPA ServicesMuleSoft Docs
• Processing Documents and Retrieving Results With the IDP APIMuleSoft Docs

More in Business & Leadership