“AI is very sexy … And it sounds really good and it's going to be helpful and solve these issues for you. But the reality is about 70% of AI products fail,” said Ron Williams, CEO of Business Evolution, during a panel discussion at the SCOPE Summit last week.
Why do so many products fail? Because they start with AI, he said. “And you can’t start with AI; AI is the last step of the process.”
The first step is fixing workflows, said Williams, “If our workflow is broken and we don’t fix it, not only are we moving bad data into AI, we’re going to accelerate the bad process.”
Combining bad data with bad processes is going to yield a “disappointing result,” he added. And the successful implementation of the technology is going to be a long road – and as Williams said, AI is not the panacea.
“Breaking down the thought process is the first change that will allow innovation to be leveraged,” he added, calling the industry’s behavior “very consistent with the old way of thinking.”
“I need you to engage a different thought process and different framework,” said Williams, providing insight from outside the industry.
Offering his thoughts as the head of clinical trial analytics at Bristol-Myers Squibb, Balazs Flink, MD, said that while pharma likes to think it is unique – it is “unique in the way that every other industry is unique.”
Flink said, “We have to constantly challenge ourselves,” to meet the need of patients and providers amid a rapidly changing value framework.
Still, the challenge of successfully using AI is not to be diminished. “AI works best in a relatively standardized, stable environment,” said Flink. “We are not there yet.”