Like most newer technologies, artificial intelligence (AI) and machine learning (ML) promises exciting ways to revolutionize and improve how things are done. In life sciences sales and marketing, that can mean predicting physician prescribing patterns, hypertargeting healthcare professionals (HCPs) with optimized and personalized messages, and delivering next-best action recommendations to sales reps.
But how much of this is actually happening within life sciences companies? Is the integration of this technology as seamless as the results it promises? Are the results even actually promising? When marketers and reps are handed AI-based recommendations, are they willing to just blindly follow them no questions asked? What kind of actual, real-life impact is AI and ML making on life sciences sales and marketing practices?
To get a better idea of the promises, pitfalls, and future of AI and ML, PM360 turned to Wesley van den Heuvel, Director of Omnichannel Engagement at Novo Nordisk and Jason Bernstein, Executive Director & Head, Medical Communications Strategy at epocrates. In the following email interview, Wesley offers an inside look into the issues and benefits that have come with AI within Novo Nordisk and Jason provides a wider view as an industry partner in this space.
PM360: What has been the biggest struggle with implementing AI within your organization?
Wesley van den Heuvel: I find the initial challenges come with determining the right areas/tactics to start utilizing AI tools and aligning with the various stakeholders as well as aligning on the right key performance indicators (KPIs)/measurement plan to track impact. I still see a lot of people who look at it with a degree of wariness in that it is replacing a person or letting a “machine” do the thinking. As an organization, we need to provide education and be clear that the technology is accelerating the ability to process and organize data into insights so that we can continue to make the best decisions to deliver valuable customer experiences.
Jason Bernstein: Wesley makes a really important point here. AI isn’t an off-the-shelf solution that can simply be allowed to “make the decision.” It takes time, ongoing adaptation via learnt experiences, and access to continuously evolving data sets based on audience behavior that requires human inventiveness to reach the next level.
When it comes to suggestions made by AI, are you seeing employees question those recommendations? How are you helping employees better understand the AI’s recommendations or decide whether it is something they should do?
Wesley van den Heuvel: There definitely can be a level of resistance to suggestions that come from an AI engine, some of it grounded in “Is this going to make my role obsolete?” all the way to, “Is a machine driving strategy?” The goal, remains, the education of the key stakeholders and providing a better understanding on how this is an important tool. If anything helps better refine strategy at a quicker pace, provides a means to pull-through brand strategy in more real time vs. a plan of action (POA) or training sessions, and ultimately is delivering a better customer experience as we are able to react to customer engagement across the ecosystem much faster with more timely and relevant information then we want to make sure we are using it. A key learning is starting with a single brand, or a subset of the sales force, and working closely with them to demonstrate the value and gain advocacy and then growing/expanding from there.
Jason Bernstein: Our role as a partner is to proactively advise our clients as much as we can on how our platform works and in turn showcase how engagement can be optimized. Beyond opens, clicks-throughs, or impressions, there is more to tell, such as: Did the behavior of the healthcare professional change? With these insights on hand, we can then guide the algorithm. At the same time, we would suggest that a balance is maintained between providing a more consumeristic or personalized experience and delivering content which carefully is disseminated at the most relevant settings.
Similar to Wesley’s comment, we have seen early stages of AI-driven programs take a cautious approach (such as short intervals of three to six months of trials and then pauses for learning). AI is one of the best tools a marketing team has at their disposal to automate processes, with an intuitive ability to digest vast volumes of data. Nonetheless, acceptance and confidence-building does take time through cycles of trials and errors, and subsequent refinement.
Which use cases for AI have actually been successful? What can you share that demonstrates that success?
Wesley van den Heuvel: We have several workstreams looking at this from smaller, easier to implement projects with the ability to demonstrate impact to gain momentum and learn. Concurrently, we are working on a parallel path of larger more enterprise-wide strategies. The absolute most important factor is education of what it is (and what it isn’t) and the potential it brings to the organization in efficiencies, customer experience, and ultimately business impact. As the organization starts seeing its role in future strategy, coupled with smaller wins such as AI-driven subject lines, optimal send time, and call to action copy…they see the potential of the larger enterprise solution.
For instance, we partnered with a company called Phrasee to work with us on AI-derived subject lines that started with two brands and now increased to the portfolio due to continued improvements in customer engagement. We are expanding into other copy elements such as pre-headers, calls to action, and website copy to see how far this capability can work in our digital ecosystem.
Jason Bernstein: One of the key ingredients of these smaller wins that demonstrate success is the data behind them. As the saying goes, “you are what you eat.” AI and machine learning are only as good as the data they consume, so any successful implementation of AI mandates clean and rationalized data ready for analysis. In some cases, this may mean upgrading from legacy data structures to a commercial data warehouse with an industry-specific data model and unified sources.
Have you tried any uses of AI that haven’t been as successful? Can you share anything about why that is the case?
Wesley van den Heuvel: As an industry, we’re navigating AI less on the software/technology side of things and must focus more on the process and regulation behind our efforts. Our partnership with Phrasee that I mentioned earlier is a good example here. Those AI-derived subject lines still must get approved by medical, legal, and regulatory (and, of course, not all of the content does receive that approval). What that means is that everything we are doing with AI has to be built, approved, and ready for us to execute in real-time and as personalized as we can get. This unique process means that, in time, it will uncover gaps in the content we have. We will need to build, approve, and load that into the algorithm and email/web templates. The AI engine must account for that, “re-learn,” and optimize. The biggest challenges the industry faces with AI is how we adapt some of our current processes and approaches to better leverage the capability.
Jason Bernstein: Additionally, while AI does a great job of taking data and analyzing it within assigned parameters, it may not necessarily be nimble enough at working outside of its lane, or even recognizing other lanes in which to work. Too often there’s an undue reliance on algorithms that leads to missed opportunities. Marketing teams should prioritize the importance of “being present” when an HCP is interacting with one of their messages in one channel, and following it up with a personalized response right at that moment, rather than vacating that space, blindly feeding in the data to the AI engine, and letting it plan the next course of action based on the HCP profiling. Even though the latter approach may be mathematically accurate, we are still losing the ability to be nimble and “seize the moment,” a strategy that is important while interacting with HCPs.
Where do you see AI going from here? How will its use within the industry grow in 2022?
Wesley van den Heuvel: I see AI-driven tools growing more and more in the next several years, both in what it can do and industry acceptance and implementation. I find the pharma industry doesn’t generally like to be first in innovations of this magnitude but doesn’t want to be last either. As more companies engage with it and see the results, I think you will see a quick acceleration of adoption. The critical piece, in my opinion, is not to do it just to say you are doing AI, but understanding what it can do for your customers aligned to corporate strategies, being thoughtful about speed and breadth of implementation, ensure stakeholder alignment at every step, and determining how it will be measured and what success looks like before you even start. I believe, like most things, you are likely to see a significant increase in AI-related projects in the industry in 2022 and with successes and failures, it will stabilize and become just part of how we do business going forward.
Jason Bernstein: I couldn’t agree more with Wesley. AI is a rapidly developing technology, and pharma is catching up fast with an exponential increase in the adoption of ML into the industry’s marketing and sales strategies. When used right, AI has the potential to revolutionize pharma marketing, especially in terms of real-time insights and predictions based on data. However, it’s not a magic bullet, and pharma should focus on harnessing the power of AI to get the marketing basics right and better engage the HCP through personalized, targeted messaging via the right channels at the right time. One thing is clear though: current adopters continue to evolve their approach as healthcare professionals are expecting the same customized storytelling journey they encounter in the digital consumer world.