Webinar Q&A: Harnessing Generative AI in Modern Automation and RPA


Thank you for joining our recent live webinar on “Harnessing Generative AI in Modern Automation and RPA.” We had a great session with insightful questions from the audience. Below are the questions asked during and before the live webinar, along with the answers from our panel members.

Webinar Recording:

For those who couldn’t attend or would like to revisit the discussions, you can watch the full webinar recording here.

Questions Asked During the Live Webinar

Is ReMark sourcing from proprietary codebases or public repositories only?

ReMark💬 uses both public and proprietary data sources. The public ones are the likes of documentation webpages and GIT repos. Robocorp has published the data loaders for the public data sources here.


What is the extent of ReMark’s knowledge of Playwright? Will it always be complete with Selenium code?

ReMark💬 can generate code in both Playwright and Selenium. Users can specify their preferences when interacting with ReMark.


Can each panel member share their views on using GenAI alongside Automation?

Rami: I recommend using GenAI with Automation for accelerating discovery projects where NLP-based AI, especially the GPT family, can be beneficial. However, I would avoid solutions that provide unfiltered results to customers or systems.

Tommi: I suggest integrating GenAI with any ongoing projects to understand its potential and be ready to implement solutions promptly.


How can we develop GenAI/Automation on current platforms but remain flexible for future platforms?

Tommi: From the start, use frameworks that decouple the models from your business logic. For example, ReMark💬Robocorp leverages LangChain and has implemented the LLM-facing parts so that the models and providers can be changed easily.

Rami: I strongly agree with Tommi on this one. For, blackboxing the AI through some (internal) API is usually a good idea, often for other AI solutions as well, with the systems developing so rapidly.


How do you address data protection concerns related to external AI platforms?

Always be cognizant of what you do with your data, both with GenAI solution and any other external platform that you send your data to. There are always tradeoffs in hosting things in-house, in the cloud, or as a Service, but especially with LLMs/chatGPT, I would use the Azure OpenAI solution while making sure my cloud architecture is safe.


How do you address the rapid pace of technology compared to a customer’s ability to adopt new tech?

Rami: This tends to be unavoidable. I personally recommend doing quick PoCs to help both customers, and yourselves understand what kind of results could be achieved and where the opportunities lie.


How does AI functionality address different error and failure rates in RPA solutions?

A major difference between RPA-based automation and AI-based intelligent automation is that with RPA solutions, we usually expect everything to be 100%  correct outside of error cases. RPA works quite algorithmically in this sense, whereas AI solutions tend to be more statistical in nature – there will always be an error rate. With GenAI, the situation can be more difficult as there are no discrete categories to check the result against, so we can’t even estimate accuracies very easily. With AI, you should always prepare to have errors, communicate to users that there will be mistakes, and plan how the system reacts to such cases.


When can we use AI without compromising our data (GDPR)?

Always be careful and check, but, e.g. Azure has various AI models available that you can use in a restricted subscription/resource group without your data escaping to third parties. Another much more complex choice is to train your own custom models, though for GPT4-level general performance, this can be prohibitively expensive.


What are the risk considerations associated with using GenAI in enterprises?

Always remember that filtering and restricting the output of GenAI language models is very hard.  Also, there is a risk of lost opportunities if you try to do everything with the newest, shiniest tool. GenAI and LLMs are great, but remember to consider if there are other possibilities as well.

Poll Q&A’s:


We appreciate the engagement and enthusiasm of our audience. Stay tuned for more webinars and discussions on the evolving landscape of AI and automation.

If you have further questions or topics you’d like us to cover, please reach out to us at info@digitalworkforce.com

New guide!

Setting Up World-Class RPA Maintenance

Why you need it and how to set it up