ABBYY hosted its AI Summit in Sydney on September 4, bringing together industry experts and thought leaders to discuss the current state and future prospects of AI in enterprise settings. The event provided a comprehensive look at the opportunities and challenges facing businesses as they navigate the rapidly evolving landscape of artificial intelligence, with a particular focus on generative AI and intelligent document processing.

One of the central themes of the summit was the significant financial commitment required for large-scale AI implementation, particularly in the realm of generative AI. Jae Park, VP of Sales - APAC at ABBYY, highlighted the enormous capital outlays being made by tech giants:

• Alphabet, Google's parent company, reported a staggering 91% increase in capital expenses in 2024, leading to a 5% drop in share price.
• Microsoft's capital expenditure for fiscal year 2024 rose by nearly 60% to $69 billion, with the company generating over $5 billion in sales from generative AI products.

Park emphasized the long-term nature of these investments, quoting Microsoft's call for patience from investors, promising returns "over 15 years and beyond." This timeframe presents a significant challenge for many enterprises, as Park noted:

"Imagine having that conversation with your board. For the next 10-15 years we need to increase CAPEX north of 50% but you can maybe possibly deliver returns in 15 years."

The summit also referenced comments from Jim Covello, Head of Equity Research at Goldman Sachs and a Partner at Sequoia Capital, who recently posed the question, "What $1 trillion problem will AI solve?" Covello expressed skepticism about the current trajectory of AI investments, stating:

"Replacing low wage jobs with tremendously costly technology is basically the polar opposite of the prior technology transitions I've witnessed in my 30 years of closely following the tech industry."

ABBYY's Approach: Purpose-Built AI

In response to these challenges, ABBYY presented its concept of "Purpose-Built AI" as a more practical and accessible approach for businesses unable to make decade-long investments in generative AI. Jae Park introduced this strategy:

"If you're like 99% of the other enterprises out there that expect returns in less than 15 years you might want to consider a different way. At ABBYY we talk about something called Purpose Built AI," said Park

ABBYY offers purpose-built models for intelligent document processing at the recently redesigned ABBYY Marketplace, a platform offering pre-trained AI models for specific document processing tasks. These models are designed to be deployed within days, achieving high straight-through processing rates out of the box. This approach aims to provide a more immediate return on investment for businesses looking to automate document-centric processes.

Ethical Concerns and Data Privacy in the AI Era

A significant portion of the summit was dedicated to discussing the ethical implications and privacy challenges associated with AI applications. Maxime Vermeir, Senior Director of AI Strategy at ABBYY, led a panel discussion that highlighted these concerns through a thought-provoking scenario:

Vermeir described a situation where a company was considering using customer service call recordings not just for traditional agent training, but to create detailed customer profiles using AI analysis.

“Now when you get that script telling you the call is being recorded for training purposes, you would typically expect the organisation will use it to train a person in customer service to talk to me better the next time.

“However, this company was looking to take this a step further, by extracting data from the recording and using it to create more in-depth profiles about their customers.

This raised important questions about data consent, privacy, and the potential for misuse of AI-derived insights.

“How long are you able to retain that conversation? Can you use that personally identifiable information to build a profile?

“This person sounds really desperate and is willing to pay anything, can I raise the price based on the AI analysis.

“If the call turns ugly and your AI determines they are a bad customer, can you deny service. What if you deny them a loan and then share that data with their cellular provider and their employer. And now their promotion is not going to go through.

“Does any of that come under the umbrella of “Using this recording for training purposes?”

“Legislation is not moving fast enough to kind of prevent these kinds of things

Clayton Peddy, Chief Information Security Officer at ABBYY, stressed the importance of transparency in data usage and storage:

"Companies need to be willing to tell their customers, this is how we're going to use the data. This is how we're going to store it. This is the objective, and this is how you can contact us to find out more about what we're doing with the information about you."

Peddy warned that scenarios involving extensive profiling and data sharing could be "a huge red flag that maybe the technologies have gone a little too far," emphasizing the need for robust consumer protection measures.

The Future of Intelligent Document Processing

Bruce Orcutt, Chief Marketing Officer at ABBYY, provided insights into the company's unique position in the Intelligent Document Processing (IDP) space. He highlighted the proliferation of AI startups in the field:

"Right now, there's over 400 vendors in the Intelligent Document Processing (IDP) space and most of them are AI startups."

Orcutt argued that many of these startups, relying primarily on Large Language Models (LLMs) trained on internet data, face significant challenges when dealing with real-world document processing tasks. He emphasized ABBYY's extensive experience in handling complex document formats and unstructured data:

"Many people want to say we're old OCR, but we're one of the only vendors in the world that can actually make that LLM accurate. Because there's never been a document that we haven't touched and seen with hundreds of billions processed through ABBYY FlexiCapture and Vantage."

Orcutt stressed the real-world challenges in document processing, such as dealing with crumpled papers, documents with handprints, or complex tables within PDFs. He positioned ABBYY's solutions as crucial for making LLMs more accurate and practical for enterprise use in these challenging scenarios.

Practical Applications and Market Demand

Maxime Vermeir also touched on the current state of AI applications in the market:

"Everybody's asking, Why is this useful? What does it actually do? Even though we have all this power, the market is still searching for how can we make this new technology practical."

He contrasted ABBYY's focused approach with broader AI initiatives:

"At ABBYY we're not trying to solve chatbots. We're not trying to solve boiling the ocean of this whole AI problem. We're trying to say, if you have a document, it has a very specific universe of data that will be contained with the addresses, names, relationships, and amounts."

“We've optimized our system to understand that context and that's all we're trying to solve.

“We have this purpose-built AI vision and strategy to transform your data into something meaningful and actionable,” said Orcutt.