Rossum CTO on AI Digital Transformation – Is It Just Hype?
Recently, Petr Baudis joined as a guest speaker on Byte Size, a podcast interviewing leaders in the digital transformation world.
Petr is co-founder and CTO of Rossum. He has a rich background in AI and worked on many open-source projects, such as one of the GIT versions, as well as working as an AI researcher for AlphaGo. On Byte Size, Petr discussed IDP – intelligent document processing – and how Rossum is working on automating document communication.
Kieran Gilmurray: What is document automation?
Petr Baudis: Documents are the arteries of the business world. We must think about the purpose of documents. Many companies work together to bring a single product to us. Manufacturing, retail, shipping and so on, and when they are working together, they exchange documents. Documents are a way of exchanging complex information for companies. Sometimes it is machine-readable .xmls, but more often than not it’s pdf documents with different layouts.
“Documents are the arteries of the business world.”
That is why automating it is so exciting. Companies are renting big offices filled with full-time employees who are issuing and processing these documents. It also slows the world down, because often there is some document sitting somewhere in a queue for weeks, which slows down the whole business process. Automating this is a way to relieve companies of all these overhead costs, but also to make the world go faster.
Kieran: What do you see as the big shift and what do you think is going to happen in the next few years?
Petr: In my opinion, there are two shifts. One is technological. I think we will see more smart AI being used as well as a shift to the cloud from on-prem solutions. And the other shift is about the paradigm we are taking. When we look at the documents, a big part of them is two entities communicating with each other, rather than it being a piece of some internal company process.
Kieran: Is AI just a bunch of hype?
Petr: Yes, sometimes it is definitely overhyped. I think it is sometimes difficult to make sense of all the claims out there. Even to us, when we were starting Rossum, after looking at websites of different vendors it seemed everything is fully automated and everyone is offering 100% accuracy. Only when we talked to a couple of people implementing this, we learned that at that point it wasn’t solved at all and that encouraged us to apply modern AI to the process.
“A good way to differentiate between what is hype and what is real is to.. see who is offering a free trial.”
That is a very good sign that the product is real and there is no solution engineer in the background fine-tuning the results instead of the AI. And you can try it for yourself first to see how it works for you.
Kieran: Is RPA where the future of automation is going, or do you see it differently?
Petr: I think RPA is a very important and powerful tool and plays a huge role in those document focused processes. The major RPA vendors also have some basic solutions of their own which capture data from unstructured documents. But truth be told, I think the purpose of RPA is to connect a lot of diverse systems together. If you are thinking about automating communication with another company, you shouldn’t be thinking about integrating different systems within a single company, but automating communication of different entities together. And in many cases it is not easy for them to adapt to you, because they may be using a completely different language – whether that is literally or in terms of different document layouts. So you need to adapt to them.
This is where I think the AI comes in, we are not talking about just the capturing of data, but a full-stack automation from receiving the documents, getting the data out, getting humans involved when needed, automating the exception handling whenever there is something wrong on the document and what happens in that case all the way to getting all the information to the target system. And at this point, RPA can be very useful. So I think it is quite complimentary.
Kieran: What do you mean by IDP exceptions and how do you handle them?
Petr: In a traditional data capture world, exceptions are anytime when a document doesn’t fit one of your templates or zones you have configured and it needs to be done by a human. In the AI world it is less serious. The AI will extract at least part of the data and the rest needs to be tweaked in the user interface. Typically, the UI is streamlined to make those tweaks easy and use them for learning and improving the AI.
Secondly, in the communication context, it is what happens when something is wrong on the document. Things go wrong all the time. You can have a wrong VAT number, VAT rate, some missing amounts..all sorts of stuff and the document is not actually right. Then the question is if there is a mistake on the receiving side, for example if the AI made a mistake, or if it is a mistake of the sender and he should make a revision. An automated communication should be able to do two-way communication, which means it can send an email to the sender with a button, letting them validate the document themselves or send a corrected document. This way, the document doesn’t sit waiting somewhere for weeks, it is resolved straight away.
Kieran: What is so different about Rossum, what makes it unique?
Petr: In my opinion, there are two things where we have a really strong opinion and really try to push forward in Rossum. One of them is a state-of-the-art AI. Many customers we talk to say they compared numerous different solutions but Rossum had the highest accuracy and that they like the speed of learning it has. We have a strongly advanced set of neural networks that can actually deal with a high variety of documents and complex documents in terms of, for example, table complexity and other things.
But what is more important is the value proposition of Rossum, which is actually quite different from a typical IDP vendor. We really focus on the use cases which are involved in the communication aspect. We are there for companies that realize their communication is document-based and that it is slowing them down, wasting time of their people, and they want to solve this issue end-to-end. Companies that are looking for a universal gateway covering a wide variety of use cases in a single place where they can train their team, solve security and compliance and all these aspects. Something that handles it end-to-end for them from reliable reception of emails, even handling PDFs and EDIs; solving communication in a single place all the way to the integration part.
Kieran: From an independent point of view, what are the key things companies should consider when they’re looking at an IDP, an intelligent document processing platform?
Petr: First of all, I don’t think Rossum is always the right answer. There are definitely cases where a different solution can give a faster value. I would always start by thinking about the process your IDP solution should be involved with. If you have documents that always look the same and there is no variety or if they are internal – they are never exchanged, then I think in those cases a template-based solution is fine and you don’t need all the AI and cloud aspects because they won’t give you much value.
“I would always start by thinking about the process your IDP solution should be involved with.”
But if you have a large variety of documents and want to do something complex with the data from them, then it’s time to think about AI. And with AI always comes the huge value proposition of cloud, because you are saving a lot of time on maintenance as well as the initial rollout because no one needs to configure each different layout with all the extra added benefits.
Kieran: Assuming someone wants to find out more about Rossum, how do they do that and get in touch with you?
Petr: The best way would be to visit our homepage where you can find out more about the product or even go to the developer page to see more technical information such as API and so on. We also have dedicated pages for people from accounts payable, logistics, and so on. I would also recommend creating a trial account, it only takes a minute and you can upload your documents and see it in action to decide on your own.
Ready to get started?
Make a quantum leap in your document processing approach. Boost accuracy and effectiveness with an AI-powered data capture solution for all documents.