Tomas Gogar, CEO and Co-Founder at Rossum: “While the world is becoming digital-first, the challenge lies in the diversity of digital formats”

We love this story from Tomas Gogar, CEO and Co-Founder of Rossum, a cloud-native Intelligent Document Processing (IDP) solution. Rossum specialises in those transactional documents every company uses – invoices, packing lists – but how do you get hold of original docs when you’re just starting out?

“At the very beginning, I literally had to walk around Prague asking for [documents] so that we could build the training data set for training our AI,” Tomas explained.

That was back in 2017. Such has been Rossum’s success in the intervening years that Tomas was named in Forbes’ 30 under 30 list, something we’re sure the Czech Technical University is proud to see. This was where he gained his Bachelor’s degree in software engineering, followed up presciently by a Masters in Artificial Intelligence.

He pursued his first startup shortly after leaving the university, and as you will soon discover has a keen interest in using advanced technology to solve practical business problems that lead to workforce inefficiencies.

Recommended reading: Four document management systems that can transform your business

Give us your elevator pitch. What does Rossum do?   

At Rossum, our mission is straightforward: enabling one employee to process a million transactional documents per year. These documents include invoices, packing lists and orders, which often come in various formats and from different channels.

Typically, completing this task would require hundreds of people, but our advanced AI technology allows for seamless automation. This smart and fast processing not only helps businesses scale efficiently but also makes them more resilient to fraud and operational risks.

What is a typical problem you are solving for clients?

Our clients frequently face a high-velocity and hugely varied deluge of transactional documents. By helping them automate these processes end-to-end, our solution drives significant time savings and increased accuracy. Our most common use cases include finance operations like accounts payable and order management, logistics and quality assurance. We serve clients across almost all continents, enabling them to become more efficient and focus on their core business functions.

Rossum has been around since 2017 – at what point did AI come into the mix? And how has it changed your offering?

Rossum has been AI-first and cloud-native since day one. Always focused on applying AI to enhance document processing, we initially harnessed it as an assistant to support human users with data capture and accuracy. Over time, we have advanced our capabilities similarly to how self-driving cars have evolved.

At first, self-driving cars were very good on well-lit highways in the US with the right (clear) road markings, but not so good when driving on poorly-paved country roads in a small Italian village. Now, further training on broader data has brought them to a point where they can successfully navigate different types of roads.

The same analogy could be made with AI. No longer limited to following a limited set of instructions, AI today is so advanced that our solution can process 70% of documents completely autonomously, including managing tasks like spotting anomalies and escalating them with senders. For example, a company’s accounts payable team might receive an invoice containing a mistyped IBAN. In addition to identifying this discrepancy, AI systems can save valuable time by instantly writing an email flagging the issue to the vendor and avoiding a mispayment.

This continuous improvement underscores our commitment to leveraging AI for greater efficiency. It’s not just about automating tasks; it’s about effectively handling increasingly difficult situations to improve and streamline human work processes. Our human-AI engagement remains robust, largely due to our investment in creating an intuitive and user-friendly interface that’s instantly accessible to anyone. By combining a powerful, continuously improving AI with a fast cloud platform and an intuitive UI, we have developed a truly exceptional solution.

Related reading: AI document capture company Nanonets earns $29 million in its quest to automate data entry

Your LinkedIn page mentions that Rossum has developed its own proprietary Transactional Large Language Model, the Rossum Aurora. How does it work and what makes it unique or special?

Rossum Aurora is our next-generation AI engine that leverages our extensive dataset of transactional documents and user interactions — in fact, it has been trained using the largest dataset of annotated transactional documents. This allows our AI to learn from billions of data points monthly, which translates to better out-of-the-box accuracy and immediate benefits for customers.

Our large language model is specifically trained for transactional documents, which differ from typical text data (think ChatGPT for example). It has been fine-tuned to operate in a unique way that guards against hallucinations by taking a discriminative approach to data capture: where standard GenAI processing tools might fabricate certain information when they can’t find it in documents, our solution only reports what is and isn’t there. Our focus on cloud infrastructure and user interface has enabled us to create a robust and reliable AI system tailored for this domain, which is prone to risks.

IDC recognised Rossum as a Leader in the Intelligent Document Processing space – what is it about it that you believe placed it above Major Players and Contenders? Conversely, where does it need to improve to take on market share from other Leaders?

Rossum stands out because we are part of the new wave of cloud-native and AI-native technology. This foundation allows us to innovate rapidly and maintain a competitive edge. To continue leading, we must stay close to our customers: making sure we understand their evolving needs and quickly develop smart solutions to address their challenges. Our journey from a startup to a leader in the field indicates our commitment to improving and adapting.

The Document Processing software space has grown hand-in-hand with the explosion in digitisation. There are now more software vendors in the space than ever. As more of the world becomes digital-first, what are some of the obstacles to growth? Similarly, what are some opportunities?

While the world is becoming digital-first, the challenge lies in the diversity of digital formats. For example, invoicing standards vary widely across countries and companies. This heterogeneity means that while most documents might be accessible in a digital format, their layouts are not standardised, which means manual processing is slow and arduous. With these problems, however, comes an opportunity for intelligent software like Rossum to automate and streamline processes. Additionally, the increasing computational power required for AI poses a challenge, but it also drives innovation and opportunities in AI development.

Worth reading: Adobe upgrades Acrobat with Firefly image creation and its AI Assistant

Shifting topics slightly – the company now has more than 100 employees and clients all over the world. How has the company culture changed over the years?

Growing from a small team to over 200 employees has certainly been challenging. In the early days, the smallness of our team inherently made for a more organic and close-knit culture. As we grew, we had to become more intentional about nurturing our culture and learn to be deliberate in how we hire, work and communicate. Despite the growth, we maintain the core principles of learning fast, working smart and believing in technology’s power to change the world.

You studied in Prague and that’s also where Rossum started. What is it about Czechia that has made it harder or easier to launch and grow a company? What are some of the strengths and weaknesses of its tech scene?

Starting in Prague presented unique challenges and advantages. One major challenge was the initial data scarcity. Funny story: at the very beginning, I literally had to walk around Prague asking for transactional documents (invoices, POs, etc) so that we could build the training data set for training our AI. One thing we immediately realised was that all of the documents were in Czech and this posed an issue because the Czech market alone is too small. So, we then focused on future-proofing the solution and enabling it to serve a broader customer base, with part of that work involving configuring our system to be adaptable for different languages and formats.

Prague’s strong pool of engineering talent was a significant advantage. Although finding go-to-market leadership locally was more challenging and necessitated a broader search, Prague’s engineering strength has been a key factor in Rossum’s success, along with our adaptability.