AiThority Interview with Don Zereski, Vice President of Engineering at Indico
Know My company
Tell us about your interaction with AI and other intelligent technologies that you work with, in your daily life.
My day usually starts around 5:30 am by asking “Alexa, what’s the weather today?” After a train ride into Boston, I swipe my Visa at a local coffee shop where the transaction is analyzed by a bank’s fraud detection system that I’d bet is powered by some form of AI. And then I spend much of my day working with our team to build AI-driven systems at Indico.
What made you Join Indico? Tell us about your role and responsibilities in the company?
The combination of a great team and opportunity got me to join Indico as VP of Engineering. My prior companies grew by solving just one or two of the many problems Indico is capable of handling. I see a much broader opportunity at Indico to automate a wide range of business processes that do not need to be human-intensive.
My role at Indico is to lead our Engineering team. Think of what we do as bundling the cutting-edge work from our Machine Learning team into a package that lets business users solve automation problems. It’s a role that bridges AI, Product Management and Systems Engineering.
How does Indico’s Process Automation exactly work around the content?
Indico Intelligent Process Automation (IPA) enables enterprises to automate business processes and workflows that depend on the analysis of unstructured content; e.g., documents, text, forms, emails, images, etc. These pieces of content are critical to the way the business operates. For the first time, this type of content can be analyzed effectively with Artificial Intelligence to automate manual, time-intensive business processes that slow cycle time and increase costs.
Indico IPA does this via a unique application of a proven Deep Learning methodology called Transfer Learning which enables the users to deploy AI to unstructured enterprise content more effectively while eliminating many of the common barriers to adoption. There are several important aspects that make it work.
First, we’ve built a huge, generalized dataset based on over 500 million pre-labeled examples – trained to understand human language. Indico IPA is able to leverage the intellectual abilities of algorithmic Deep Learning models to make accurate judgments based on the information and context available vs. hard-coded business rules or expert systems. This allows users to train their own, custom models with up to 1000x less data than required with other AI and Machine Learning solutions.
Indico IPA also provides “Explainability” so users understand exactly what data inputs impact the decision outputs. This makes it easier to tweak models to ensure accuracy and for Data Science and line of business, teams to collaborate based on a common understanding of how it works.
The Indico IPA interface also enables SMEs to be more involved in defining use cases and good training data and training Machine Learning models. This helps Data Scientists and line of business professionals to set realistic expectations for their initiatives. The benefits are significant up to 85% faster cycle times; up to a 4X increase in organizational capacity and throughput, and the ability to redeploy valuable resources to higher-value activities for the business.
What is your view on the global trend of including ‘AI’ in everything’ into businesses?
There is some reality to that. AI is increasingly a part of how we do things in business and in our personal lives. In our experience, AI is really no longer “what” customers are buying. There are far fewer companies today, vs. say two years ago, looking for “AI” solutions. Instead, they’re looking for business solutions and having AI can be an important and differentiating feature in that solution.
I think you’ll see fewer AI-only solutions coming to market, and see fewer pure AI startups being funded.
How do you customize the Indico’s products according to job role(s)?
Indico’s product isn’t custom-tailored to the user as much as our users create custom solutions with it. We have a wide range of users from Ph.D.’s to analyze using the product to solve real business problems using pipelines of models made with Indico IPA.
Which industries can best leverage Indico’s content process automation?
To date, we’ve seen the most traction in Financial Services Banking, Institutional Investment, insurance. Part of that is simply a matter of focus on our part. Financial services companies depend on the types of document-based workflows I mentioned earlier to run their business. These workflows are mission-critical to how they serve their customers and make operational decisions. Commonly used cases tend to center around back-office processes related to Legal and Compliance, Sales and Support, and Finance and Operations.
Realistically, any business that depends on large amounts of unstructured content and document-based workflows can gain a lot of benefits. If it requires a “judgment,” you need the intelligence platform like ours.
What is that one recent instance wherein Indico’s offerings helped an enterprise solve a critical business problem?
One client is a major US bank that processes over 400,000 institutional requisitions for funds disbursement annually. Clients send instruction documents to the bank to disburse wire payment to third parties. The existing process involved (human) processors manually checking the bank’s systems to ensure funds were available and comparing the trust agreement specifications against the instructions of each wire, the authorized signers’ list, as well as the correct entry of the wire in the disbursement system. Each set of instructions might be in a different format, order, and structure. As such, it was a very time-intensive and manual process, but customers counted on the bank to get it right every time.
Leveraging Indico’s IPA platform, the bank was able to automate the extraction of the required information from the varied client instructions received in inbound pdf and emails and integrate that into the bank’s wire disbursement platform. As a result, the bank is projecting annual gross savings of $2.4 million and an 82% reduction in processing cycle times for each transaction.
How do you gain information on AI/ML and related topics?
It’s tough these days because there is so much being generated, but industry sources like AiThority and Information Management are really helpful. And wsj.com has created a section devoted entirely to AI which I find very helpful in regards to the practical application of the technology on the business side.
What are the difficulties in the deployment of AI?
There are many reasons. Some are inherent to the complexity of the technology and some are self-inflicted by users. But there are a few common things we see. In too many cases, AI is a solution looking for a problem to solve. Users have to start with a business outcome in mind. Otherwise, they risk completely wasting their time.
Some users are unrealistic about what AI can and cannot do especially with unstructured content. AI will not tell you the answer to something you don’t know in a large pool of data. But it is great at discovering what maps to something you’ve defined as the desired state. You have to be able to define that desired state for AI to add real value. But that’s harder to do with unstructured content.
Access to the right data is probably the most common obstacle we see. You don’t necessarily need a lot of it, but you need to have very clear examples that define what is right vs. wrong. Then you can put AI to work against a large dataset. When it comes to data, we advise clients to look internally for that data vs. externally. Scraping data from the internet is generally not very effective.
Data Science teams going it alone is another common misstep. Related to my comment about starting with a business outcome in mind, Data Science teams can only go so far without the subject matter experts in the business. We encourage customers to team their Data Science pros with a line of business from the start. This helps drive them towards good use cases with a measurable ROI.
Where do you see AI/Machine Learning and other smart technologies heading beyond 2025?
The technology is advancing so fast these days that it’s really hard to look that far out. But I’m betting on a few macro trends that I think will have a big impact.
AI and smart tech will be embedded in most of the everyday products and services we use in our business and personal lives. As part of that, I think privacy considerations will take on a much more integral role in those interactions. Companies that allow customers to easily personalize that aspect of how much data they share about themselves will be at a considerable advantage.
We will see the power shift from Data Scientists to the line of business with these technologies. You already are seeing AI-based solution providers putting more capabilities and tools into the hands of line of business subject matter experts and other non-technical folks. Whether you like the term “citizen data scientist” or not, making the tech easier to use is essential to maximizing its value.
Explainability will continue to increase in importance as regulators and key stakeholders put greater emphasis on knowing why/how their AI algorithms make the decisions.
People are starting to understand that AI is really more about augmentation vs. replacement through pure automation. There are certain tasks that can be fully automated and tools like RPA will continue to gain traction for those used cases. But the real opportunity in AI is with higher value tasks where AI is a tool to make people smarter and more efficient so they can spend their time on higher-value activities for their organizations.
What AI start-ups and labs are you keenly following?
What technologies within AI and computing are you interested in beyond the current work?
Beyond Indico, I’ve been interested in the evolution of AI hardware. For things that directly touch what Indico does and what I see our customers doing, I’m interested in Chaos Engineering and watching the enterprise shift to Cloud-native apps.
What’s your smartest work-related shortcut or productivity hack?
There’s a great story in “Good Strategy, Bad Strategy” by Richard Rumelt that involves Andrew Carnegie paying $10k (a ton of money in 1890) for a piece of sage advice. The story goes that Mr. Carnegie was at a cocktail party where he was introduced to Fredric Taylor. Mr. Taylor was on his way to fame for organizing work to be more efficient. Carnegie posed a challenge… “Young man, if you can tell me something about management that is worth hearing then I’ll send you a check for $10,000”. Taylor paused then offered this bit of wisdom… “Mr. Carnegie, I would advise you to make a list of the 10 most important things you can do. And then start by doing number one”.
It’s timeless advice and highly relevant to the noise and distraction-filled world where we all live and work. Take the time to understand what’s most important to your customers, team, career, family, etc. and spend your time working on the first item.
Tag the one person in the industry whose answers to these questions you would love to read.
I’d love to hear from my former HERE colleague, Hans Peter Brondmo, now at Google.
Thank you, Don! That was fun and hope to see you back on AiThority soon.
Don Zereski has always been at the forefront of technology. From early internet start-up companies like Tripod to the HERE Division of Nokia.
Don has excelled at product strategy and building the platforms and applications that monetize that vision. Prior to joining Indico to lead its product development efforts, Don held Senior Executive roles at oNotes, HERE – a Nokia company, MetaCarta, a pioneering enterprise local search company and Lycos, where he was responsible for revenue and product strategy. Don holds an MS in Engineering from WPI.
Indico is the leading provider of Intelligent Process Automation (IPA) solutions. We help organizations turn the process into profit by enabling them to automate manual, labor-intensive, document-based workflows. Our breakthrough in solving these challenges is an approach known as transfer learning, which allows users to train machine learning models with orders of magnitude fewer data than required by traditional rule-based techniques. With Indico, enterprises are now able to deploy AI to unstructured content challenges more effectively while eliminating many of the common barriers to adoption.