VOLGOGRAD, RUSSIA - SEPTEMBER 28, 2020: A thermal imaging screen shows children's body temperatures ... [+] at the entrance to secondary school No 17 in Voroshilovsky District. Dmitry Rogulin/TASS (Photo by Dmitry RogulinTASS via Getty Images)

Frost & Sullivans new September 28, 2020 report,The Reshaping of Industries Caused by COVID-19, identifies nine global trends that will emerge as a response to Covid-19. The analysis says that pandemic preparedness will speed the deployment of AI solutions and accelerate AI innovation and rely on AI and machine learning (ML) tools to expedite digital transformation across crucial business initiatives.

But, McKinsey Digital says that Covid-19 has allowed them to see the Achilles heel of artificial intelligence (AI) and analytics.

While Covid-19 has been a catalyst for AI and analytics adoption by many businesses that hadnt digitized their operations, Tamim Saleh, Senior Partner, McKinsey Analytics, says the predictive powers of analytics are failing.

While analytics has helped many businesses to navigate the crisis, it has also shown fallibility, said Saleh. As a technology that relies on the details of past patterns and behaviors to predict what will happen next and what action a business ought to take, Covid-19 has thrown a bit of a spanner in the works, given this predictive model is no longer effective, with the pandemic having transformed the way we live, and work.

With the data feeding into algorithmic models having been built before the crisis, a lot of predictive models have lost their relevance, breaking down as a result of Covid-19, said Saleh.

Saleh believes that this is because these models are no longer reflective of the world we all live in today. In the same way, we didnt have a pandemic handbook telling us how to adapt; they too, dont have the data to predict what will happen next, adds Saleh.

According to Saleh, the retail sector is a good example of irrelevant models.

In retail, models were often built primarily on physical footfall. This [..] stopped when the pandemic struck. Retailers werent anticipating all their sales to move to online channels and this is not just a short-term change, said Saleh. Our research has shown that around 60% of consumers who are new to digital channels have decided to stick at it. Following the Covid-19 pandemic and the shift in consumer behavior, this had meant that when the physical stores reopened, the algorithms for forecasting, budgeting and revenue instantly became outdated.

To get to more accurate predictions again, the only solution is to re-engineer models that are relevant for the world we live in today, said Saleh.

Now, more than ever, businesses need to be pragmatic to survive, said Saleh. In the next normal, there are too many unknowns, and so human judgment is critical in all decision making. Today, for many models, it is not possible to get the perfect data or historical correlations that can help with predicting future events. This is where human judgment comes in.

Saleh says that businesses need to embrace the 80:20 rule, which means changing the entire analytics approach.

Instead of building teams and plans based purely on analytics and data, it is time to re-prioritize human judgment, said Saleh. At the beginning of 2020, predictive modeling and AI judgment were sophisticated enough for many businesses to rely on. Today, you need to do so with caution and realize that the human element has become, in many ways, more important.

Saleh believes that there are several ways to categorize disruption to analytics and AI, starting with the level of sophistication of the model and how features are weighed.

The more complex your model is, the more opaque it is, so it is more difficult to gauge when its predictive powers may be in question, said Saleh. Looking at theaters, while a lot of data exists around ticket bookings, pricing, etc., we cannot [..] say in any way shape or form, how they will operate in four years from now.

However, on the other side of the coin, theres the nature of the model itself. If your AI and analytics model is being used to measure quality control in a factory, then likely, that model will not have been heavily disrupted by Covid-19, said Saleh

But Saleh points out that if youre using AI and analytics models in industries highly impacted by human behavior, such as retail, then Covid-19 will have caused a lot of disruption to the models in place.

This is because ultimately, the pandemic has impacted our lives in many dimensions such as what we buy, how much we spend, and how we react to advertising messages, said Saleh. Therefore, any model relying on analytics around behavior can very quickly become obsolete. And so, to recover, we need to be pragmatic.

Read the original:
McKinsey Digital Says Covid-19 Has Revealed The Achilles Heel Of Analytics - Forbes