4.1.19
Today Aible launched the industry’s only Real World AI trained for business impact, not accuracy. Aible lets business users create custom AI based on real cost-benefit tradeoffs and operational constraints. This innovation shows business users the expected business impact of the AI before deploying it. Aible makes AI accessible to anyone with knowledge of their business, no data science experience required.
“All AI to date has been trained on the wrong thing — accuracy,” said Arijit Sengupta, founder and CEO of Aible. “Businesses need impact, not accuracy. Aible is turning AI on its head with a breakthrough that empowers business users to make impactful decisions for their specific challenges. We call this Real World AI.”
According to Sengupta, today’s AI — with its focus on model accuracy — is broken. Accuracy doesn’t lead to impact because AI trained on accuracy doesn’t take into account that costs of different types of errors are not equal. To understand the importance of business impact over accuracy, consider this sales example: What if the benefit of winning a deal is 100 times the cost of pursuing a deal? You might be willing to pursue and lose 99 deals for a single win. An AI that finds only one win in 100 tries would be very inaccurate but would boost your net revenue.
Tim Darling, founder and chief analytics officer at Laudio, said, “The translation between the business owner and the data analyst takes a lot of time in my experience, and Aible shortcuts most of that communication chasm. For example, out-of-the-box machine learning models optimize on accuracy, which gives equal weight to the cost of failing to predict something vs. incorrectly predicting something will happen. There are ways to adjust for this – you can manually adjust thresholds after a model has been run or you can run a model with a penalty matrix. But both of those require a lot of mental gymnastics and accurately communicating business needs to the analysts running the model. It takes time and clear communication. There are many other steps in the process of setting up and iterating models that take weeks. Aible talks to business owners in the ROI and constraints language they use and understand, reducing the back-and-forth discussions. In our case, we predict which nurse might quit their job so a hospital can intervene early and try to retain that nurse. The cost of failing to predict a nurse quitting is very different than the cost of unnecessarily intervening with a nurse who would have never quit their job. Hospitals have limited resources to follow up with nurses to retain them. Building cost-benefit tradeoffs and constraints into the models that reflect the business realities of the organization is hard to do; Aible does this automatically and gives me optimal recommendations we can act upon.”
“When we think about AI, we can’t just focus on simple metrics of AI quality. We need to start with business outcomes if the AI is going to have a business impact,” said Gregory La Blanc, distinguished teaching fellow at the Haas School at UC Berkeley. “I will use Aible’s Real World AI in my ‘Data Science and Data Strategy’ MBA class this semester to help students understand that the business impact of AI matters more than measures like accuracy.”
