July 24, 2018/Lauren Schivley/Blog, Decision Support, Omnichannel

Q&A with Veeva’s Paul Shawah and Aktana’s David Ehrlich – Part 2

As discussed in part one of this series, machine learning and AI have already been applied in several ways in the commercial life sciences — yet there’s potential to do even more. In part two, Veeva Systems SVP of Commercial Strategy Paul Shawah and Aktana CEO David Ehrlich discuss what’s in store for this emerging AI and machine learning technology and how commercial life science and medical teams can make the most of it.

Ten years from now, how will commercial teams be using AI? Will AI replace the sales rep?

DE: This is one of my favorite questions. For the near to medium-term, we really cannot remove humans from the picture. AI systems are great at sifting through data and finding patterns, but they fundamentally can’t understand why those patterns occurred. We need humans to look at those insights and make the final decision on which to use.

Consider chess grandmaster Garry Kasparov’s experiment on the benefit of human-machine symbiosis. After being beaten by IBM’s Big Blue, Garry put together a PC-based chess program that provides suggestions to mid-ranked chess players. When pitted against the chess grandmasters, the aided mid-ranked chess players consistently won. This isn’t surprising — there are some things that machines do well (like finding patterns in data) and some things that humans do well (like judgment calls or understanding other people). When you combine the two, you get better results.

PS: David spoke about the human judgment element, and that is critical to making AI even more powerful. The other aspect is about human relationships. HCPs make important decisions about patients’ healthcare every day. In such a complex environment, it’s not always as binary as accessing the right data. It is about service and relationships. For example, you can get all the technical and product information you need about a car online. However, most of us still rely on a salesperson at a dealership because they provide service and a relationship to help us make an important decision. Combining digital capabilities like AI with the human element — the judgment and service of the rep — leads to stronger, engaged HCP relationships.

We’ve talked about commercial life science applications a lot. What does this look like on the medical side? Where will machine learning have the biggest impact here?

PS: Medical Affairs also struggles with technology adoption, often because of its relative size compared to commercial and the investment requirement that goes along with it. But the opportunity for machine learning is just as big, if not bigger, than for commercial applications. Similar to commercial, the medical industry can use machine learning to better understand the customer and identify the scientific information their key opinion leaders (KOLs) want and when. The unique challenge on the medical side is the data itself, which is significantly more complex. Determining the best engagement plans for key opinion leaders will require scouring through broad insights from publications, presentations from congresses, and clinical trial participation — as well as unstructured data sets from labs and real-world evidence. The opportunity to use AI to do so is great.

DE: That increased complexity is also reflected in the many sub-questions that need to be asked on the medical side, such as who is your key opinion leader in a particular therapeutic area and in which geography are they influential. Once the key opinion leader is identified, you have to figure out how to best communicate with them, which message to send and through which channel, and how to keep them as an advocate if they are indeed interested.

This gets back to the whole notion that no matter how good you are, you’re going to need partners to truly succeed at using machine learning for relationship-building. The opportunity for machine learning is too big for one company or medical team to take on by themselves.

We’ve seen similar application of AI in online retail and other sectors for 10 years or more, and yet the life sciences industry is just getting out of the starting gate. Why is the life sciences industry late to adopt this technology?

PS: It’s taken so long simply because the industry didn’t have a compelling reason to change for so many years. The commercial model used to be very simple — sales reps sold to HCPs that worked for themselves and made their own decisions. You had more reps, you reached more prescribers, you made more sales. In the last five to seven years, however, the model has changed — HCPs are becoming employees of medical groups, access to prescribers is decreasing, and there is price pressure on drugs. With increased pressure to win over these HCPs, life sciences companies are quickly adopting machine learning as a way to create more effective customer experiences.

While they are a little late to the party, I think they will move fast with machine learning. If there’s a compelling need for technology (as there is now for machine learning), life sciences companies will quickly adopt it. The iPad, which was introduced to pharmaceutical companies in 2010, was rolled out to most companies within just a couple of years. If it had been introduced 20 years ago, it wouldn’t have caught on because there wasn’t an urgent need for it.

DE: The life sciences industry faces more challenges than other industries when adopting new technologies also because it is highly regulated. With multiple FDA approvals, HIPAA regulations, and confidential doctor and patient data to protect, companies intentionally and understandably approach new technologies carefully.

As life sciences companies are adopting this technology, many now have multiple AI outputs. How can companies ensure they aren’t creating more noise and that they all work together and harmonize as one?

PS: It’s typical of emerging technology to produce multiple separate outputs, because it’s hard to do everything all at once. The industry is in early stages of widespread adoption. Companies are implementing AI for a particular sales team or therapeutic area – perhaps they don’t yet have an AI initiative on the marketing side. The logical next step is they’ll work towards optimizing at this individual level and then scale or merge across areas. But it’s hard to optimize when you’re looking at it from an individual level and these data sets are separate. We’re seeing companies starting to bring these data sets together and consolidate — I think there will be more of this going forward.

DE: Having a clear idea of what life sciences companies want to achieve with AI will help them stay on track, particularly when they design AI-infused processes that reflect those objectives. If you’re trying to get human beings to make decisions in a reasonable amount of time, you have to give them the knowledge they need in a very concise and directed way – whether it’s insight derived from a standard channel or through AI. Ultimately, AI itself can synthesize and find the signal within the noise across intelligence applications.

If you missed it, read part one of the series, where David Ehrlich and Paul Shawah discuss the current state of machine learning in the commercial life sciences and how it’s changing the HCP experience for the better.