Work Ahead Webinar Series: AI-Enhanced Automations – Combining Transformative Capabilities

Frequently Asked Questions


  1. What is AI?
    • Artificial Intelligence (AI) is a branch of computer science that enables development of intelligent machines, like ML models, that function and react to dynamic situations like humans.
  2. What is ML model?
    • Machine Learning (ML) models are created based on the data sets provided for the machine to learn how to react to different data in dynamic situations. ML models observe data and its processing to become trained for specific tasks.
  3. How do RPA and AI work together?
    • Robotic Process Automation (RPA) is used to automate routine business processes – both back-end and front-end – which are usually highly manual and consume plenty of time and costs. Still, many business processes are more complex and cognitive, thus, they can be automated only with a combination of RPA + AI. Owing to AI skills, the robots can better understand how to deal with different types and formats of data in accordance with specific business needs.

AI Fabric

  1. What is UiPath AI Fabric?
    • AI Fabric is an end-to-end platform that helps accelerate AI & Automation strategy. It enables customers to easily deploy their own AI skills, pre-trained AI skills developed by UiPath or UiPath AI technology partners. You can drag and drop them to an RPA workflow via UiPath Studio. Monitor and manage your AI skills, constantly train and improve them using the human validated data.
  2. What are the use cases for AI Fabric?
    • AI enables automation of processes that include:
      • Uncertainty – Processes in which the outcome cannot be determined with 100% certainty (i.e. property valuation, loan defaults, inventory forecasts, etc.).
      • High Variability – Processes that have too much variability and many dependencies (i.e. resume matching, purchase decisions, language translation, etc.).
      • Unstructured Data – AI enables you to deal with information like articles, documents, images, videos, and emails.
    • With AI, your robots are able to learn new skills. AI-enhanced robots can read, write, listen, interpret, and make decisions. This helps companies automate more complex and cognitive processes bringing their automations to a new level.
  3. Can I add AI to the existing projects?
    • Yes - UiPath AI Fabric enables you to instantly apply the power of AI to any software currently being automated by RPA.
  4. What are the key features of AI Fabric?
    • Model of Your Choice - We don’t limit you in terms of model choice. The open platform provides out-of-the-box models and supports models built by you and UiPath partners.
    • Drag and Drop AI - Choose a model and drag and drop it to your workflow right in Studio to accelerate the potential of your automations. Select an AI skill suited to your business needs from a drop down.
    • Model Governance - Gain end-to-end visibility of model versioning and updates. Your RPA teams will know when models have been updated, and your data science teams will know how your models are being used in production and will have traceability on what data they were trained on.
    • Continuous learning enabled by Human-in-the-Loop - Continuously improve your models by bringing Human-in-the-Loop to handle exceptions, validate data, and send it for retraining.
  5. Can I use my own ML models?
    • Yes - Package your models, upload them directly to AI Fabric, and make them available for robots to consume within minutes.
  6. What out-of-the-box (OOB) models do you provide? Can they be customized according to my business needs?
    • UiPath provides a range of pre-trained ready-to-use packages: Sentiment Analysis, Question Answering, Language Identification, English To French, English To German, and English To Russian. Also, UiPath’s Document Understanding models (which include retrainable out-of-the-box models for receipts extraction, invoice extraction, and a general document extraction) will be easily deployable through AI Fabric. Many others are under development now.
  7. How do I train ML models?
    • If the prediction a model has made is of a low confidence level (encoded in the RPA workflow, the exact confidence threshold should be decided by the Data Scientist working with the RPA Developer and other SMEs), the platform brings humans to validate results and feed the data back to the model to train it. Although model training can still happen outside of AI Fabric and can be imported as a new version of a model, in the next release AI Fabric will enable a seamless experience for model retraining within the platform. UiPath AI Fabric enables version control and the ability to revert to a previous version.
  8. Does AI Fabric support training of my ML models or is it limited to those provided by UiPath?
    • The same rules apply to all models (your own ML models, pre-trained ML models developed by UiPath or UiPath AI technology partners). Each model can be retrained to ensure the highest accuracy.
  9. Is AI Fabric already available? Is it a part of the UiPath Free Trial?
    • Yes - AI Fabric Cloud is in General Availability. The on-premises version will become available later this year. The product is available for free as part of the Enterprise Cloud Trial with certain limitations.
  10. Do you have an example of a real use case with results presented?
    • With AI Fabric, Australia’s largest mutual bank improved the accuracy of living expense assessment classification, resulting in:
      • Increasing automation potential from 40-50% to 90%
      • Saving countless hours of tedious labor
      • Reducing the amount of backend work
      • Improving customer and employee experience

Document Understanding

  1. What is UiPath Document Understanding?
    • UiPath Document Understanding is a new product offering that combines RPA and AI to help you extract and interpret data from different documents and ensure end-to-end document processing. The tool works with a wide range of documents from structured to unstructured, recognizes different objects like tables, handwriting, signatures, or checkboxes, and can deal with various file formats.
  2. Is Document Understanding similar to OCR/ICR?
    • No, it’s different as it can “understand” documents - extract and interpret data - not just “read” them. Yet Document Understanding uses OCR as a technology for digitizing documents, that is reading the text.
  3. What is included in UiPath Document Understanding?
    • Document Understanding framework is set up in Studio using the IntelligentOCR activities pack (which will be renamed to Document Understanding soon). It includes all activities needed for end-to-end document processing. For the data extraction itself, Document Understanding offers several extractors relying on rules, templates, and ML models as extraction methods. Correspondingly, you will get access to pre-trained ML models and to the model training/retraining capabilities via AI Fabric platform. Finally, Validation Station within the Action Center is used to confirm extracted data and handle exceptions if needed.
  4. Which types of documents can be processed with Document Understanding?
    • Structured documents like different forms, passports, licenses can be processed with rule-based and template-based extractors. For semi-structured documents like invoices, receipts, purchase orders, we recommend dedicated ML models. Most importantly, for any types of documents, a hybrid approach can be applied to ensure the highest accuracy. This means that you can combine various extractors even for a single page or field. You can use a hybrid approach to process unstructured documents as well or opt for partner integrations – like Indico or others – who offer leading technology on the market for unstructured documents.
  5. What are the pre-trained ML models available in Document Understanding? Have you created your own models, or do you support the existing models provided by other vendors?
    • Pre-trained ML models for Invoices and Receipts developed by UiPath are entering General Availability on 5/4/2020. Our ML models for purchase orders and utility bills are currently in preview, and many others under development. You can also retrain these models on your custom data to ensure the highest accuracy. Similarly, you can create/use your own ML models or those provided by partners.
  6. Can Document Understanding process handwriting?
    • Yes – you can use Intelligent Form Extractor for handwriting processing.
  7. Which OCR engine is most accurate?
    • There is no definite answer to this question as the result will be different for different documents and OCR engines. The only way to define which one is better for your specific needs is to try them out. UiPath Document Understanding offers several OCR engines of your choice, including Omnipage OCR, Microsoft OCR, Google OCR, ABBYY OCR, Tessaract OCR, etc. Also, our native UiPath OCR will become available soon.
  8. How does Validation Station work? Does the model improve based on the corrected data? Is this step mandatory, or can we set a certain confidence level allowing to bypass this step?
    • Validation Station is used by an employee to confirm extracted data, if the minimal confidence score isn’t reached, and to handle exceptions. Thus, if the minimal confidence score is reached, this stage is bypassed. As for model retraining based on the human validation, these capabilities will be released in Q3 2020.
  9. Is Document Understanding already available? Is it a part of the UiPath Free Trial?
    • As part of the 2020 FTS Release, UiPath Document Understanding Cloud enters General Availability and UiPath Document Understanding On-Prem enters Limited Availability. The product is available for free as part of the Community Cloud and Enterprise Cloud Trial with certain limitations.
  10. How is the product licensed?
    • UiPath Document Understanding offers consumption-based licenses for 16,000-page bundles.

Action Center

  1. What is offered as part of UiPath Action Center today?
    • The main feature offered as part of UiPath Action Center today is Actions. This is the centralized inbox accessible through web-based portal with specific privilege or access control for users to take the required actions. In this release we provide Actions as part of the Orchestrator. The Robot creates a task and suspends the workflow. Business Users can access this task from the Actions inbox and take the necessary action
  2. What are the other major capabilities of UiPath Action Center?
    • Apart from Actions, you get to use Long Running Workflow (LRW) that is now built into the Studio as a set of new activities. Long Running Workflow allows for automating end-to-end business processes and linking of multiple processes. With LRW you can orchestrate the Robot and Human tasks. Process monitoring is also offered with UiPath Action Center. This helps monitor the long running workflow from start to finish as a single job spanning across different business units. You can now get process execution summary snapshots of Robots, humans and triggers to identify and resolve resource bottlenecks.
  3. Is Uipath Action Center available on cloud?
    • UiPath Action Center is available both on Cloud and on-premises. Currently the community release version is available on cloud only.
  4. Can we access UiPath Action Center through mobile phone?
    • Yes. You can download and install the UiPath Mobile App from App Store and Play Store for tour iOS or Android devices. The Actions of UiPath Action Center can be accessed and managed from the mobile app. Users can manage or take the necessary actions such as approval, validation, etc from the mobile app from anywhere at any time.
  5. Is Validation Station available for the user when using Action Center?
    • Yes, Validation Station is integrated with Action Center. With the Validation Station integration, users can easily handle document exceptions. User can just drag and drop the create document validation activity into the long-running workflow. Users need not create a new form, instead they can directly provide the inputs for the required fields available through the Validation Station and they get to see the image of the document right beside the form.

AI Computer Vision

  1. What technology enables AI Computer Vision?
    • AI Computer Vision is powered by a custom developed visual neural network, specifically designed for use with RPA.
  2. What is the cost to use AI Computer Vision in my automations?
    • AI Computer Vision is currently free, available for use with all UiPath Robots.
  3. What are some of the primary uses for AI Computer Vision?
    • Typical uses for AI Computer Vision in the creation of automations inside of Virtual Desktop Interfaces such as Citrix, and for interfaces that experience minor changes over time.
  4. Is AI Computer Vision only on-premise?
    • AI Computer Vision is available both on-premise and on the cloud.
  5. What hardware requirements are required for on-premise deployments of the AI Computer Vision server?


Q: I’m keen to know more about the Document Processing Framework, especially around PDF files.

A: Sure! Find what you need in our online documentation for UiPath.PDF.Activities and UiPath.DocumentProcessing.

Q: Is Computer Vision able to track handwritten text present in a document? Or does this still require Abby Flexicapture?

A: UiPath Computer Vision focuses on reading UI (User Interface) elements on the screen with AI techniques. We have several Cognitive services integrated with the UiPath Platform, including Microsoft Azure Computer Vision and Google Cloud Vision, which can recognize handwritten text.

Q: Is Computer Vision recording related to the recent package which has the privacy statement that screenshots may be sent to UiPath?

A: Yes. There are, however, two deployment modes for this feature. If you access it through our Cloud, the screenshots are sent to UiPath. But you can also deploy the model on-premises or in your own Cloud. In that case, nothing gets sent to UiPath.

Q: Does your AI Computer Vision require a license? Is Computer Vision available in the community version? The community edition currently sends images to the UiPath server. Will a robot do that after the release of the enterprise version?

A: Computer Vision is available in two ways today. First, there is a Cloud-based option available for free to our community and enterprise users. This is the default method configured in the activity. Second, the model is available for on-premises use by enterprise customers. In the case of the Cloud option, images are sent to the UiPath Cloud for processing, but UiPath does not retain any of the processed images. For the on-premises option, images are not sent to UiPath at all.

Q: With the Computer Vision activities, can you anchor to a variable string? I have a grid of text boxes with various row/column headers and would rather use a simple loop to go through the anchors to find the text box I need to enter data into.

A: Yes, you can do that.

Q: What is the maximum duration for Orchestrator Execution Snapshot recordings before an unattended robot encounters an error?

A: Two minutes.

Q: Is there any capability for the integration with ServiceNow API in which we must pass credentials?

A: We are actively exploring the integration with ServiceNow and will have more to share in the future.

Q: Regarding monitoring: is that information available via an API query? Is there any historical data available?

A: The information to query the data available in monitoring is not available via API. Monitoring data happens over 24 hours. For longer durations in monitoring, you can ingest the logs via elastic-search to compute utilization/uptimes.

Q: Can I create a Global Variable in my Main.xaml and access it via sub-workflows without using Arguments?

A: Global variables are usually considered a bad practice and are not supported.

Q: How is UiPath related to RPA and classic Artificial Intelligence areas (as i.e. neural networks, rules-based systems, expert systems, knowledge-based systems and so on)?

A: Here’s our own vision for AI.

Q: How can I get involved in the Fast Track 2019 program?

A: The 2019 Fast Track is released and available for download. If you wish to participate in future platform testing, you can join the insider program.