- In the wake of the coronavirus pandemic, companies retrained their machine learning models as a response to unusual customer behavior, according to Vikram Mahidhar, senior vice president of digital at professional services firm Genpact, speaking at an MIT Sloan CIO Digital Learning Series panel Wednesday.
- The pandemic has led to consumer shifts in behavior, which gave businesses a chance to build more empathetic responses into their automation strategies, said Mahidhar. For example, in the financial sector, companies can improve customer's experience by taking into account economic, social and health challenges.
- Retraining algorithms is a start. "For the foreseeable future, we'll have to continuously go back to these models and refresh them until we settle on them," he said. "For some companies we're doing this on a weekly basis," in response to shifts in data and behavior.
Customer and user behavior follows patterns. Personalizing or accelerating workflows according to these trends is how artificial intelligence and ML deliver value for businesses.
Systems rely on past data to deliver projections about the future, such as the likelihood of a person stocking up on groceries in bulk monthly purchases or staggered trips to the supermarket.
The effects of the pandemic and stay-at-home orders reduced the efficacy of many assumptions about consumer behavior. For at least 13.3% of Americans, those effects include navigating life without a job.
Taking consumer shifts into account can result in more effective results. For example, banks looking to execute on debt collections would previously have had "dialers" calling customers non-stop until they got through. But now the world is changing, Mahidhar said.
With ML models, banks can predict when a particular consumer has a propensity to pay debts back in 60 days instead of 30 days. Providing customers with a self-pay plan via email gets "a lot more response in certain segments" than calling, Mahidhar said.