Interview with Joe Dumoulin, CTO & Innovation Officer at Verint Intelligent Self-Service

Joe-Dumoulin-CUE-CARD



At Verint, we see machine learning as the training tools that power AI. In other words, machine learning algorithms are continuously evolving with better data and in turn powering customer engagement AI.

Tell us about your journey into Machine Learning and how you started at Verint?

I became interested in machine learning while working on ERP and production control systems in the early 90s. For about seven years, I had a job installing computer systems in factories and working directly with production staff. Learning how they solved production planning problems was an incredibly eye-opening experience. I was able to apply many interesting techniques to these problems and learned a great deal about the overlaps between decision processing, control theory, and optimization. Machine learning sort of appears in the confluence of these ideas.

How do you prepare for an AI-driven world in a tech-heavy marketing industry?

Marketing has always been quick to adapt and adopt when it comes to technology. There’s a risk tolerance and value in experimentation within marketing that many other parts of the organization are more sensitive to. That’s why marketing has stayed at the forefront of AI adoption, starting with automated ad buys and moving into customer engagement and new voice and NLP technologies.

How do you differentiate between Machine Learning and AI at Verint?

At Verint, we see machine learning as the training tools that power AI. In other words, machine learning algorithms are continuously evolving with better data and in turn powering customer engagement AI.

What are the foundational tenets of your data mining and machine learning technology? How do you make MarTech and AI-tech platforms more powerful?

First and foremost is the business bottom line. Data and machine learning simply for innovation’s sake is not useful for the enterprise. We focus very intentionally in first steps on specific business problems and solutions and believe that aligning our solutions with these specific business goals gives the platforms the true ROI and impact that MarTech needs to show results.

To what extent are machine learning output supervised? Can they be tamed and/or manipulated by humans to tune performance further?

Absolutely, and from our perspective, human training is essential. While machine learning can certainly operate and evolve in a vacuum, our decades of experience in developing ML models and AI engagements have proven that human oversight and input is essential to help guide the effectiveness of those programs.

The extent to which supervision is needed will vary project by project, but the human touch is always a valuable component of the process and development.

What gaps do you commonly find in the applications of AI/ML? How do you fill those gaps at Verint?

The biggest challenges remain problems of context and intent. We’ve focused on solving those problems by developing deep domain expertise for each solution. We’ve produced the largest library of business intents that any provider can offer, which allows us to bridge that gap and continue to expand understanding and actions for our solutions.

How do you train your teams to manage ML for better adoption in the industry?

Our research team is very, very focused on applying machine learning to areas that are substantive for our customers. Sometimes ML and AI appear to be a solution looking for a problem. We take care to focus first on business problems or goals, next on bringing our skills to play to solve those problems. Often these are subtle issues that come up because of the inexact nature of AI tech.

For example, when should the AI escalate a conversation to a live person?  We have developed some powerful tech around this problem to accurately determine when a customer is likely to need to speak to an operator before they ask for it. This is a recent, but very promising project from one of our researchers.

What are your predictions on the role of AI for digital communications?

In a decade, AI will be the backbone of all digital communications for the enterprise. Customers are coming to expect automated solutions to engagements and service problems, and while human agents will obviously still be involved, AI will be at the frontlines of those interactions.

Tell us more about your recent work on AI and machine learning technology at Verint?

I mentioned escalation prediction. Some other recent research initiatives are focused on mixing heterogeneous classifiers, rule-based and statistical, to get better performance, researching how humans establish grounding and relationship in conversation and how to train systems to understand human relational language, using attention (as in RNNs) and linear programming as methods for surfacing why complex neural network models make the decisions they do. These initiatives have all been slowly moving through our research team.

We are gearing up on a number of other exciting initiatives that involve active semi-supervised learning tools, discovering anomalies in text data, and trying to discover emergent trends in text data streams.

How do you make AI and machine learning deliver economic benefits as well as social goodwill?

At the very outset, AI should be deployed with a focus on ROI and the bottom line and the economic good of the company. These business objectives should be at the heart of every AI deployment and be built into a customized solution that can speak to an enterprise’s unique needs.

As far as how companies deploy that AI in a manner of social good, we believe that it’s important that technologies that touch customers reflect brand values and the company’s dedication to customer experience.

What are the major challenges for intelligent mobile technology companies in making it more accessible to local communities? How do you overcome these challenges?

Mobile has actually opened up access to more local communities than any other technology we’ve had before. That’s why it’s so crucial that we develop AI that can integrate and deploy effectively with mobile platforms so that these advances can continue to reach local communities. After all, AI technologies are at their best when localized and personalized, and this is the opportunity that mobile provides.

Tell us the “Good, the Bad, and the Ugly” of AI/ML technologies – How do you prepare for those disruptions?

The Good: The automation that AI and ML offer will transform productivity and even the capabilities of what we can achieve. The downside of that is that companies need to prepare their workforces to understand, work with, and optimize these technologies, rather than think of automation as a replacement for human employees. We’re not seeing enough preparation for this from companies right now.

The Ugly is that we are at an evolution point for AI and machine learning where the way we interact with this technology has plateaued, and we need to rethink and begin developing for an AI 2.0 enterprise environment. We have constructed AI technologies as call and response, and the next generation will need to be more intuitive and focused on context and action.

The Crystal Gaze

What AI start-ups and labs are you keenly following?

I love the work coming out of the Rasa team. I think that enterprise customers want to be able to keep their data local in ways that are difficult with most of the tools offered to be only available in the Cloud. Rasa is just getting started, but I am looking to see some major advances from that team in the near future.

Read More: Solving The Security Problem Means Solving The Human Problem

What technologies within AI and computing are you interested in?

Our team is very interested in using machine learning to enhance rule-based language understanding. We believe that businesses need to understand why decisions are made and need to incrementally influence how understanding happens in their domain. We’ve proven this over and over in the past decade-and-a-half. For that reason, we are very interested in bayesian methods and probabilistic programming applied to these types of business problems. Along with that, we have a strong interest in tooling that makes it possible for business team users to directly train and control deployed systems without needing full-on data science resources.

As an AI leader, what industries you think would be fastest to adopting AI/ML with smooth efficiency? What are the new emerging markets for AI technology markets?

Industries that are data-driven and data heavy, like the financial sector, are the most naturally primed for AI deployments, which is how we’ve seen the market develop. As for where AI will move next, the biggest efforts are behind healthcare and transportation, but these advances are as much about data as AI.

What this history show is that data is the predictor for AI adoption, and really the first step needed for AI to transform a market and help an industry reach its full potential.

Recommended: New And Exciting Ways Brands Are Using AI On Social Media

What’s your smartest work related shortcut or productivity hack?

I like to teach at the local university to keep in touch with what students are learning and to find promising new talent for our team. I also find that spending an hour a day in the gym really helps me be more productive.

Tag the one person in the industry whose answers to these questions you would love to read:

Frankly, I am more interested in our customer’s and our vertical’s opinions than industry, although senior management at Microsoft really seems to get the idea of business first, then technology.

Thank you, Joe! That was fun and hope to see you back on AiThority soon.

Leave a Reply

Your email address will not be published.

You may use these <abbr title="HyperText Markup Language">HTML</abbr> tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

*