How Novartis Procurement is Cutting Failed R&D Experiments with AI

In the latest episode of the PharmaSource podcast we spoke to  Mark Burfoot, Vice President Research Procurement at Novartis about how advances in Machine Learning allow procurement to have a more strategic role in optimising life science research.

Prior to his work at Novartis and Pfizer, Mark’s educational background is in immunology. That knowledge of the scientific process has allowed him to create new frameworks and forms of value that go far beyond cost savings.

“There are different ways of approaching procurement.” Mark says. “You can take the traditional and transactional approach –  build contracts, help your scientists gain access to the array of suppliers and services that they require in order to achieve their work.”

“But with the advent of machine learning techniques and other types of approaches, there are other ways of looking at a problem that can help redefine how we can add value. 

Rather than looking at the cheapest product, or trying to provide the widest range of options to a scientist, we can now offer tools that help the scientist to choose the products and services most likely to lead to success.“

Optimising R&D

The insight procurement provides has been proven to reduce the number of failed experiments, putting Mark and his team in a far more strategic position to steer and inform decisions about which projects to invest in.

“The nature of research is that there is a huge amount of experimentation and failed experiments. But a lot of experimental failure happens just because we haven’t picked the best reagent.”

“We are exploring tools and approaches that can help us  guide the scientists and give them insights to say, ‘“’Hey, you’re better choosing this set of reagents versus testing all these other ones that most have have a cost effective approach.'”

“There’s a cost gain to that, but it has a productivity impact because now you’re moving faster and getting less failed experiments” as a result of that. “

Mark and his team are now helping the scientists make more informed choices which are proven to increase the success rate of experiments.

“Suppliers will always tell you their product is the best and will meet all your needs. And as I know from when  I was in the lab, scientists are still very dependent on reading materials and methods and what appears in peer reviewed literature.

Procurement is now able to provide a new perspective, offering insights to suggest better ways of doing things – different materials and methods which are more likely to get the end result they are looking for.    

“It has become a very unique partnership. We are redefining the ecosystem and the supplier base and the reagents and the services that are required in order to help drive the science.“

Measuring success

“For me, driving the speed of the science more effectively is so much more valuable than just giving more choices and having the best contracts in place. “ say Mark.

“Adoption of some of the tools that we use in this space have had a 70-80% uptake, across our scientists globally. There is still a massive opportunity in this space to do more, but the desire to play, to adopt is very good.”

“By making these kind of improvements it’s possible to see savings of 10 -20 million USD over a period of a few years in a particular region or class, which you can use for reinvestment in other areas. You can keep exponentially growing.”

“We’ve seen significant increases in productivity as a result of improvements in different types of reagent classes that we purchase, for example. That has a very positive effect. I would say that there’s probably been a four-fold increase in time saved.”

“You’re getting both cost savings on one end, but you’re also getting speed and productivity gains at the other end that’s helping you accelerate the science. It is a really nice position to be in.”

“Rather than just time saved, I want to understand what’s the output of productivity saving. Is it that the number of failed experiments is reduced? “

The next step is to take these new metrics of success back to their suppliers. “They’re not used to it either, but if we hit these new metrics of success then we will buy more of their product or service”.

Change management

Mark explains that for this type of approach to be a success, it needs to be accepted and welcomed by the research business. 

“If you pitch this as “we’re replacing scientists with AI” that’s probably not a particularly powerful way to achieve change management and actually, that’s not the intent.”

A better approach is to present stakeholders with the new options they can choose from, so that they are still making the final decisions and still in control.

“They can choose to ignore it or to use it, but you’re giving them data and statistics and a level of analytics, that’s helping them make informed choices. People tend to react positively if they can see and trust where the data is coming from and if the results coming out have led to a positive outcome.”