Home Bots & Business AI, the next step in automation

AI, the next step in automation

AI Summit addresses challenges in effectively scaling automation

by Marco van der Hoeven

CIOs and CFOs who want to maximize the benefits of automation increasingly turn to the next step, the addition of Artificial Intelligence (AI) functions. Increasingly, automation solutions are designed to help organizations tackle not only single processes but achieve end-to-end automations, by adding  technologies like machine learning (ML) and natural language processing. This week UiPath organized the AI Summit, where these trends were discussed. Rocking Robots sat down with Boris Krumrey, Global VP Automation Innovations at UiPath.

According to Boris Krumrey, Global VP Automation Innovations at UiPath, customers are currently certainly very interested in AI. But there is a catch. “The difficulty for a large organization is: how do you make it applicable to the actual processes on the ground and the business operations? You have to deploy it, maintain it, operate it. If you look at the amount of machine learning models that are really in production, it’s a small percentage compared to all the models people are creating. So what we still see a lot in business is narrow AI models.”

This means for example specific kinds of prediction or classification models, “If customers want to do a prediction around their cash flow, we create for them in the UiPath Immersion Lab the initial model that they can train, and then drive forward and optimize. With this knowledge they can put actions in place to improve cash flow. Many of our customers are very interested in those kinds of models, because departments like finance, HR and IT all started with discovering robotic process automation (RPA) first, and now their interest is shifting onto AI.”


“The benefit of our platform is that it already has an AI engine included. So you can upload any kind of machine learning model. These prediction models can be integrated just like an input-output model into the RPA workflow. That makes it immensely powerful.”

Another area which attracts a lot of interest is document understanding and intelligent document processing. “It started with document capture, which is now combined with RPA. The models we have use machine learning, so it’s not a static model.  You get the new document, you do the corrections, then the corrections feed into the next run of the auto ML-loop to train and improve the model. This is proper machine learning, not a template based solution.”

End-to-end automation

“The nice thing is, you can use it for example to start with the invoicing model, and then modify it and use it for totally different documents. You can even use it for documents that include handwriting, unstructured messages and email. All this interaction is done end-to-end by robots. We see a huge demand for this at the moment. And more new and interesting solutions, like computer vision models, are coming.”

He mentions a solution with a technology partner that has developed a machine learning model for detection of items on the shelf in a retail setting, connected to the UiPath platform. The robots are taking live streams from cameras on the shop floor to see whether an item is out of stock on the shop floor. The system automatically  sends a notification to resupply the shelves.

Customer experience

Right now most AI-related demand is centered on document understanding and intelligent document processing. Natural language processing is another area a lot of organizations are interested in. “Companies have now realized that these are technologies that really help them improve the customer experience. We also see a high demand for combining AI capabilities and process understanding. So already our task mining-product is using a machine learning algorithm to identify the best opportunities for automating.”

The level of maturity to really leverage AI is already there in the market. “Most companies have already started somewhere with their automation. And I’m talking to more and more companies who are scaling now. They’ve done the implementation, and have successfully proven the benefits of automation. Now they want to go further. To do that they have to do more than look at processes. Eventually you don’t want to automate processes, you want to automate organizations. Our mission is broader than just the process.”

Executive support is key to a successful implementation

The maturity is also reflected in the level of executives responsible for automation. ”We definitely talk to C-level about automation. The challenge most CFOs have, for example, is how to save costs in the commodity standard shared services area, and reinvest that money into an area that is closer to the business. That can be manufacturing, or customer service. But they all have to digitize the operation. And we are talking more and more to CIOs, who now are becoming closer to the business operations.”

This was reflected at the AI summit. “We address AI in particular scenarios of particular industries. We’ll take away the mystique around AI, it’s really about building blocks. You don’t need to be a math expert, or a data scientist. That gives you a lot more areas where you can apply machine learning. We take a very practical approach for people to understand how AI can be applied.”

“The key takeaway is to understand how machine learning models, combined and embedded with automation, can really help you in your current daily operations, whether it’s document understanding, computer vision, unstructured text extraction, natural language process discovery,”.

Watch the recording of the AI Summit here



Misschien vind je deze berichten ook interessant