Clarkston Consulting’s Mike Onore recently attended PharmaUSA 2023, a two-day event bringing together pharma leaders and colleagues to discuss emerging topics and critical issues across the industry – from becoming more analytics-driven organizations, to the need for stronger data strategies, to emerging technologies. He shares his key takeaways below.

PharmaUSA 2023 Recap

1. The Nature and Role of AI in Pharma 

At PharmaUSA 2023, the biggest difference between the discussion on stage versus private conversations was the nature and role of Artificial Intelligence (AI). On stage, presenters spoke of great advancements in becoming more analytics-driven organizations, with predictive and prescriptive modeling having taken huge leaps to become part of standard business decision-making at many firms. However, a second wave of innovation is coming. Generative AI and Large Language Models (LLMs) have the potential to again reconfigure what it means to be a data-driven organization.  

Business leaders are recognizing the realities and challenges associated with the unending promise of AI. At Pharma USA 2023, it was clear leaders were amid a reconciliation between how far we’ve come and an evolving, unsettled future – which I’ve organized into “The Five Stages of AI Grief:”

1. Denial: At first this appears to be a tricky one. So many solution providers promise AI, and most pharma companies mention it as part of their strategy, so where is the denial? It shows itself in the denial of certain inevitabilities of AI. There’s still a popular (and warranted) trope about generative AIs “hallucinations,” but we’ve already seen with search incorporation, limited training sets, and other guardrails how these are already being addressed. AI will be less and less “wrong” every day. Another element of denial is that what we’ve built in the industry may largely have to be reworked around advanced in LLMs. LLMs (like ChatGPT) will be the dominant form of AI in our lives, but it’s not even part of the vernacular yet (during at least one panel, no panelists were familiar with the acronym).  

2. Anger: The anger is at the black box. Pharma leaders like control, they like agency, and they like an explanation. All are fading. A panel moderator announced, “Human touch will always be important,” but next year I might be hearing, “Human touch will always be important, right?” We’ve seen OpenAI and its 375 employees leap over Google and its 15,000 employees, largely on the backs of Microsoft’s giant, parallel, supercomputers. I don’t believe any jobs are being automated away, but the nature of work will change, and we will have to overcome resentment during that transition.  

• • There’s a certain irony here. A long-standing trend is Home Office encroaching on field decision-making, with data, dashboards, and pushing compliance with omnichannel alignment. A speaker even said that the rep should no longer own the relationship. What happened to the sales rep as the quarterback of the healthcare provider (HCP) relationship? Home Office owns data-backed insight generation, if not full decision-making control. But now, commercial insights teams will face their own specter – black-box AI – impinging on their recently acquired domain. The lesson here is about adaptation over fear.  

3. Bargaining: This is the healthiest relationship we can have with AI. What can the machine do better than me? What can I do better? What conclusions am I willing to accept without fully understanding the machinations that developed them? Where are our business rules helping, and where are they hurting? We might also want to ask” 

• • Do I want advanced analytics as part of metadata management (see knowledge graph-based data catalogs)?  
  • Do I want it on the front end? (see natural language self-service query and visualization tools) 
  • How much experimentation in prescriptive analytics can we stomach (see Next-Best-Action frameworks)?  
  • How long of a leash does the robot dog get when it’s barking and snapping at my customers? 

4. Depression: The depression is from broken promises of “out-of-the-box” solutions. When executing in a business context, we face the hurdles of any new technology. Often, integration of productionization of models was more challenging than expected, or we wanted to build something where the tech wasn’t mature enough to support it. Often, business partners may be uninterested or unwilling to accept conclusions of data science teams, a new-ish seat at the commercial and medical table. Maybe we picked the wrong use case, building a sandcastle with a bulldozer or visa-versa. There is (fortunately!) a lot of hard work left to do. A question from the audience showed how we’ve hit inevitable roadblocks: “How can NBA (Next Best Action) be less reliant on email to deliver at scale?” The answer is complicated, and AI won’t be the magic wand. 

5. Acceptance: The crucial final stage is built on trust. During the panel, I thought about how often I got to the second page on Google. The trust is built over so much time and hard work. Epidemic SEO and sponsored results have only now started to erode that trust. I thought about our patients and getting them to put pills and injections into their body without fully understanding science. We’ve built that trust with providing as much transparency as possible, by academic rigor and consistent outcomes. A speaker commented that as we move from predictive to perspective analytics, the change management will come from how well we articulate the positive effects. The journey of building trust can and will happen in careful hands. 

2. Proliferation of Real-World Data

There is no shortage in data types and data quantity for pharma companies. Years after even the term “Big Data” has become passe, most data assets aren’t fully leveraged. The first asset type, omnichannel data, is established around better capture in customer relationship management (CRM) and mature digital channels. Pieces like paid search and rep-triggered email are already standard. Now, partnerships across commercial are increasingly focusing on Real World Data (RWD), outside of traditional Real-World Evidence (RWE) or health economics and outcomes research (HEOR) teams. Merging omnichannel with claims, electronical health and medical records, and other RWD creates a fuller picture of the HCP. Pairing all that with relationships/connections within a knowledge graph creates the full 360-customer view. Access to RWD is increasing too. The rise of synthetic data will complete the picture of patient cohorts, and federated data networks will provide access to new clinical trial data while maintaining regulatory adherence.  

3. Data Strategy as Foundation for Impactful Analysis

Presentations on successful projects explicitly called out focus on data as a product to wrangle data proliferation. A few stories centered on successful Customer Data Platforms (CDPs), with foundations of common data models that are flexible and robust. Layers of embedded analytics and feature enrichments lead to predictive models and automated insights in production. Another presentation was dedicated entirely to the data lifecycle – strategy, acquisition, and management. The focus on upskilling marketers and ops teams will not just be tools and self-service analytics, but on data governance. The speaker correctly noted data and governance teams will have to demonstrate the goal isn’t bureaucracy, but a foundation to derive real and consistent value from organizational data.  

PharmaUSA 2023 Recap: Looking Ahead

At PharmaUSA 2023 there was no debate on our goals as an industry – to serve our patients and our customers better. However, it was a fantastic forum to engage with differing perspectives on how to achieve those goals. As the new data landscape crystallizes, successful implementation will require a willingness to learn, experiment, and adapt. To chat more about the conference or emerging trends across pharma, connect with us today.