AI on Demand: Data Science in Operations

Reading Time: 3 minutes

Artificial intelligence (AI) is right here, right now—and it’s changing our lives. The ever-present need for business optimization, combined with a long history of applied statistics, explosive growth in available data and recent advances in cloud computing has accelerated innovation and business transformation.

However, creating and implementing AI systems is tricky – many things to get right, many technology options and many human and business considerations. This article covers the major functions and issues of AI at a high level. The accompanying webinar and series of workshops show how to run AI systems at scale, and how to enable AI-driven applications across a variety of functions and business areas. The workshops feature TIBCO Connected Intelligence technology, which embodies all the major functions of AI – at scale in a single platform.

AI technologies

In recent years we’ve been teaching machines ever-more human tasks. Speech recognition and in-home agents like Amazon’s Alexa. Image recognition like Clear identifying and checking travelers in at the airport with fingerprint and eye recognition. Natural language processing and chat-bots that help customers troubleshoot their cable service. These AI techniques are firmly rooted in some areas, but still difficult to get AI to function on-demand and get Data Science into operations. “The future is here, but it’s not evenly distributed” as William GIbson says.

Some AI-driven systems we take for granted as part of daily life. For example,  recommendation systems analyze at our current behavior and purchase history and make personalized suggestions — Amazon for online purchases, Netflix for movie recommendations, and Spotify for music playlists.

More than this, the many real-life business applications that I see with TIBCO customers, is a constant reminder of the value driven by AI and Data Science. AI helps all industries perform tasks otherwise not possible — such as managing financial risk, spotting fraudulent transactions, detecting and treating diseases, optimizing energy production, detecting anomalies in the manufacturing of computer chips, forecasting demand, engaging customers and protecting the environment.

So, where did all this AI stuff come from all of a sudden?

Well, it turns out that the core ideas of AI are based on technologies built over years of mathematical statistics and computer science, that can now be run quickly and at scale. AI is algorithms — trained on historical data and managed in computer software pipelines, generating predictions that are actioned by business rules on event streams.  

Machine Learning and Data Science

The poster child for AI in recent years is machine learning (ML), especially the emergence of deep learning. Machine learning models are trained on historical data and predict from new observations. Broadly speaking, supervised learning models predict a target variable from other variables while unsupervised learning models identify patterns in data without focusing on a target. Classification models classify new observations into various categories like whether a credit card transaction is fraudulent or not, or if regression models can identify anomalies in IoT systems such as oil and gas production.

Extreme value is generated, and businesses are transformed, when ML models and AI apps are deployed into operations — making predictions on the data that flow through a business. That state of AI on Demand with Data Science Models in Operations is the hallmark of a successful AI business system.

While AI systems are generally developed to drive innovation and business transformation, we note that there is also the potential for unintended consequence and even a dark side as described by the tehno-sociologist Zynep Tufekci in her TED Talk, “building a dystopia to make people click on ads”. Indeed, with concepts such as hyper-personalization,  people now getting the news that they want to see, and some studies show falsehoods spread six times faster than truisms. As a practical matter, we are seeing regulations and ethical oversights driving “model fairness” – preventing discrimination against age, sex, race, communities; via enforcing data and models that don’t discriminate.

TIBCO Connected Intelligence — AI on Demand: Data Science in Operations

In TIBCO’s “AI on Demand with Data Science Models in Operations” series, we highlight some of these AI success areas and showcase the underlying data science and application creation. This 10-city workshop tour features signature software demonstrations in recommenders and suggestion engines, customer engagement, dynamic pricing, risk management, and energy production surveillance. These demonstrations are based on real-world TIBCO deployments, driving extreme value for TIBCO customers across many industry sectors and functions.

We’ll be showing the new TIBCO Spotfire X and TIBCO Data Science products, and you’ll see demonstrations of how to detect anomalies in data with lots of dimensions, how to handle extremely wide datasets, how visual analytics are auto-suggested from the data, how streaming and static data sources interplay in data discovery, and how to build models and deploy them as a service.

The series features data science design patterns that we are publishing to the TIBCO Community site along the way.

Watch the webinar today to learn how you can implement AI driven by data science in your own organization.