In this Q&A with MIT/SMR Connections, Gavin Day, Senior Vice President of Technology at SAS, shares real-life examples of artificial intelligence (AI) at work, discusses picking the right problems to solve with AI, dispels a common misconception about AI, and defines AI success.
Want to read more about AI success?
Q: Could you describe some especially interesting AI use cases?
Day: Two major truck manufacturers use sensor data and SAS AI solutions to predict maintenance issues and prevent unplanned downtime, which takes a tremendous toll on the fleet operators and customers that are expecting these deliveries. They monitor the data from each truck if something is wrong with a vehicle’s major systems, such as the engine or transmission, they can take it out of service before it breaks down on the road somewhere.
Another customer, a major aerospace manufacturing company, predicts potential failure of airplane parts before they fail. But they’re also using it to see where they need to have parts distributed around the world. That’s because knowing something is going to fail is one thing — having a part ready and available where these planes are in flight is the second part.
Then there’s an organization focused on supporting healthy bee populations. They provide video footage from inside the hives, and the machine learning algorithms decode bee movement so teams can better understand where bees are finding food. This real-time monitoring of bee movement allows beekeepers to establish hives in optimal locations to maintain strong colonies.
Read more from this series with MIT/SMR Connections
One last example is that machine learning and AI are showing great promise in advising analysts when to review and make manual overrides to their forecasts in the financial industry. We’re in testing with a large global consumer goods company, and the approach has reduced the number of forecasts needing manual review, which cut analysts’ time in half and improved overall forecast accuracy by 6%.
AI’s real value comes from making better decisions that lead to better business outcomes. If we make better decisions at every level in an organization, from tactical to strategic, and we make those better decisions every single day, that’s a strategic advantage.
Q: How can organizations pick the right problems to solve with AI?
Day: They need to start with understanding and doing an honest assessment of their AI maturity and skills. If they’re just starting out, they need a project that has a limited scope and that will benefit from a single AI capability such as machine learning or conversational AI. If they’re advanced in their maturity and have AI skills, they can choose to tackle more complex projects that bring multiple AI technologies together.
Specifically, “next best” recommendation engine capability is becoming required now for marketing and sales. So, that’s a place for some organizations to start.
Q: What kinds of problems should organizations avoid tackling with AI?
Day: My initial advice: Don’t solve a problem you don’t have. Some organizations will read articles about how helpful AI is and then decide to start solving problems that aren’t core to their business. So, you need to make sure — as with the adoption of any technologies, but for AI in particular — that it’s solving a problem you need to solve.
Q: We’re hearing that some companies think they need to build AI from scratch — but, in fact, they may already have AI capability without realizing it. Could you speak to that?
Day: There’s definitely merit to this thought. If an AI application is extremely specialized, then a custom approach is the way to go. But, for the large majority of those types of tasks, we have found incorporating AI capabilities, object recognition, and conversational AI into your existing tools and workflows is the right approach.
There’s often discussion of AI being the bespoke application sitting “over there.” The SAS approach is different. We want AI capabilities in everything we do. We use the saying “Sometimes it’s hidden in plain sight.” The capabilities could already be present in what you have, even if you don’t realize it.
As an example, we have customers using SAS technologies to detect fraudulent transactions in real time. We’re a very common solution in the market, and we’re using machine learning and deep learning to improve fraud detection and find new threats. But sometimes, that capability isn’t transparent or recognizable to our customers.
Q: What does AI success look like?
Day: As we put technology into the market and it evolves, how do we know that it’s actually solving anything? In my opinion, AI’s real value comes from making better decisions that lead to better business outcomes. If we make better decisions at every level in an organization, from tactical to strategic, and we make those better decisions every single day, that’s a strategic advantage. That’s where I hope the use of AI and analytics continues, with the tide moving toward using both to become smarter companies and make better decisions — not just using technology because it is the latest and greatest thing.
