2. Executive Summary
• RPA is the marriage of screen-scraping technologies with business process
automation.
• Marketing sprinkled AI on top because, well, AI is cool.
• RPA can reducing errors and costs related to swivel-chair integration
activities
• The more back-office swivel-chair tasks you have, the stronger the ROI
• This includes QA automation activities
• For the geek, it’s:
• Transactional integration without ETL; and
• Application interaction with out API needed
3. RPA Sweet Spot
• Use-Case Examples
• QA automation
• Data validation and/or integration between
Salesforce & SAP
• Form scraping from legacy mainframe
application
• Transaction duplication between systems
Swivel chair
integration
tasks
Thousands of
common
repeated
tasks
Tasks need to be
completed quickly
and predictably
Sweet
Spot
4. RPA Types
• Attended – does an action when a user interacts with it, or tells it to
• Front office activities such as helping answer a customer inquiry
• Unattended – does an action on a batch basis with no human start
• Nightly data verification between Salesforce and SAP
5. RPA Is…
• For reducing time, costs and
errors related to automating
manual swivel chair integration
tasks.
• Able to scale to enterprise use
RPA Isn’t…
• Augmented intelligence
assistant. You can use RPA to
help make useful chatbots, but
RPA is not a chatbot.
• Going to solve unstructured data
integration problems, but could
be a platform to help
6. RPA – How it’s Positioned by Vendors
• Save time & money by automating mundane process activities
• Integrate legacy systems that don’t have an API
• Built for the cubical worker to create and a developer to tune
• Use AI to go from “taught automation” to “learned automation”
7. RPA – How it Works
• Application User
• Builds “bots” by recording initial task operations(s) by recording themselves
working with applications
• Literally operates at the application UI layer.
• Developer
• Augments bot recording by editing recorded tasks by adding operations such
as
• Data validation; Wait/delay commands
• Can build routines & abstraction as necessary using robust bespoke scripting
interfaces
• Publishes bots using management facilities from RPA vendor
8. RPA – Why Customers Buy It
• Automate slow and error prone “swivel chair integration tasks”
• Solutions are generally enterprise ready – manage & deploy
thousands of bots with failover and monitoring and security
• Hope of solving ancient but still critical unstructured data integration
problems
9. RPA & AI
• The RPA AI story is about being able to:
• Identify relevant unstructured data within content (invoice number) and knowing
what to do with it
• Generally referred to as “text analytics” For example the system finds a string of characters,
in a PDF, understands it’s an account number and links the PDF to the customer record
• Identify the application on the screen, the content being shown, and knowing what
to do with it
• Generally referred to as “image analytics or “convolutional neural network.” For example the
system recognizes there’s an SAP invoicing window on the sales reps screen and pro-actively
shows the customer record in Salesforce
• Reality
• Fear-of-missing-out has vendors talking big but reality is they are tightly controlled
demos for marketing purposes
• The theory is sound, the reality is fraught with the same challenges as any other AI
strategy