Robotic Process Automation

How do we bridge the gap between RPA and Machine Learning for Intelligent Automation?

LI Asked by Linda Martinez · 18-05-2024
0 upvotes 15,851 views 0 comments
The question

Our RPA bots are great for repetitive tasks, but they break the moment a UI changes or an invoice comes in a new format. How are people integrating Machine Learning (Document AI) into their UiPath or Blue Prism workflows to make them "smart" enough to handle unstructured data?

 

3 answers

0
AM
Answered on 20-06-2024

This is the transition from "Standard RPA" to "Intelligent Automation." Instead of hard-coding the bot to look at specific XPaths or coordinates, you use a Machine Learning model (like a Vision Transformer or an OCR-based LLM) to "read" the document. In my current project, we use an ML model to extract data from varying invoice layouts and then pass that structured JSON back to the RPA bot to enter into SAP. This way, the bot doesn't care if the "Total Amount" is at the top or the bottom of the page. You should look into "Human-in-the-loop" (HITL) configurations where the bot flags anything with a low confidence score for a human to review, which helps train the model over time. 

0
MA
Answered on 23-06-2024

Are you finding the licensing costs for these "AI Units" in platforms like UiPath to be prohibitive? It seems like adding intelligence doubles the cost of the bot. Is the efficiency gain really high enough to cover that? 

DA 26-06-2024

Matthew, the cost is higher, but the maintenance is much lower. A standard bot that breaks every two weeks costs more in developer time to fix than the AI licensing fees. In our case, the "Intelligent" bot has an uptime of 98% compared to the 70% we had with a rules-based bot. When you factor in the reduction in manual "fallback" work, the ROI is actually better with the AI-integrated version.

0
MA
Answered on 29-06-2024

Start with "Process Mining." Use it to find which parts of your workflow actually need AI versus which ones just need better, more stable rules. Don't over-engineer simple tasks. 

LI 01-07-2024

Great advice, Mary. We found that 40% of our "complex" issues were actually just bad data entry at the source. Fixing that was cheaper than building a complex ML model!

Share your thoughts

Your email address will not be published. Required fields are marked (*)

Professional Counselling Session

Still have questions?
Schedule a free counselling session

Our experts are ready to help you with any questions about courses, admissions, or career paths. Get personalized guidance from industry professionals.

Request a Call Back

Search Online

We Accept

We Accept

Follow Us

"PMI®", "PMBOK®", "PMP®", "CAPM®" and "PMI-ACP®" are registered marks of the Project Management Institute, Inc. | "CSM", "CST" are Registered Trade Marks of The Scrum Alliance, USA. | COBIT® is a trademark of ISACA® registered in the United States and other countries.

Book Free Session