When singer-songwriter Randy Travis survived a near-fatal stroke in 2013, he was left with numerous physical challenges, the cruelest of which was limited singing and speaking ability. Yet earlier this year, his voice could be heard in the new song “Where That Came From.” How? A form of artificial intelligence (AI) specialized in voice cloning used Travis’ voice, as recorded over a decades-long career, to fill in the vocals of the song. This inspiring example shows how AI can help individuals with health conditions make up for deficits they experience.
But the role of AI in healthcare will go beyond this. For years, our colleagues in R&D have successfully implemented AI to discover new molecules, simulate biological processes, and determine optimal clinical trial design. And there’s more to come. As Scott Galloway said so poignantly at a keynote in May: U.S. healthcare, with its 17% share of GDP, is “the most disruptable industry in the world,” and “the first big carcass that AI is going to start to feed on.” It remains to be seen if AI will make healthcare either less expensive or better at improving people’s health, but there is no doubt that it will be a major part of how the field will evolve.
Enabling Precision Medicine
Every stakeholder is going to be impacted. Patients can look forward to electronic assistance that can return their voice or other lost abilities. For example, Whispp is an AI application that can convert whispered or affected speech into the speaker’s natural, original voice. As we enter into the age of precision medicine, AI will be a critical tool to determine, with the help of disease knowledge graphs, which indications, associations, contraindications, side effects, genetic markers, drug targets, drug carriers, and interactions to take into account on a case-by-case basis. One step in this direction is DrugGPT, a tool that has been rolled out to doctors in the UK. It provides guidance and acts as a safety net for prescribers. When entering a patient’s conditions into the chatbot, it responds with a list of recommended drugs. It also flags possible side effects and interactions between drugs. It’s easy to imagine that prescribers are going to rely more and more on such tools.
Currently, AI appears to be capable of automating tasks but not entire fields of work. AI isn’t coming for your job. But the person who knows how to apply AI in your field may well try to make themselves comfortable at your desk. So, don’t be afraid but stay on top of your field. For example, the medical establishment was for a while abuzz with the idea that radiologists might no longer be needed if an AI can perform just as well at interpreting radiographic imagery. It turns out that we still need humans to collaborate with other HCPs and contribute to decision-making. While the field has benefited from technological advances in the isolated task of identifying abnormalities, radiologists now just work faster, more efficiently, and more reliably. But they are still needed to quality control and oversee the tech at work, to ensure accountability and responsibility for diagnostic decisions, to evaluate cases holistically, and to collaborate.
Efficiencies with AI
As for healthcare marketers, there are a number of ways AI is starting to optimize our workflows. For example, ClaimsTM, Omnicom Health Group’s compliance orchestration solution, is designed to be a brand’s “single source of truth.” In other words, ClaimsTM provides creators access to real-time approved brand claims, references, and safety during initial creative development and actively surveys draft materials throughout the project lifecycle to ensure they align with previously approved content. This promises to accelerate the medical, legal, and regulatory (MLR) process, obtain approvals quicker, and reduce rework cycles. In the patient engagement space, the potential is immense. Yet from a risk-mitigation perspective, we anticipate a cautious, measured approach by most biopharma companies. The simpler types of AI that rely on approved data libraries or thoroughly vetted scripted user interactions will likely dominate. Most bearers of regulatory responsibility are going to be uncomfortable allowing a generative AI to directly interact with patients. There have already been examples of generative AI chatbots making commitments they’re not authorized to make—Air Canada’s recent experience with having to offer a discount when there wasn’t supposed to be one illustrates this nicely. While the fallout of this rather amusing example is easily contained, a “hallucination” like this, occurring between a pharma chatbot and a member of the public, could easily have major regulatory and reputational repercussions. All this is to say: The less predictable the tool is, the less likely it will be to obtain approval.
A Bonus for Authenticity
As biopharma professionals begin to routinely implement AI solutions, one thing we should all watch out for is how unique our work outputs are really going to be. If we all use the same models that are trained on the same sets of data, all we’ll see is a race to the middle. True innovation, true thinking outside the box, cannot happen that way. For the foreseeable future, outstanding work will be done only by humans working in conjunction with AI.
The question of authenticity ties in with this: In a world where everything you see, read, and hear is potentially artificially generated, people are likely (or hopefully) becoming more savvy and discerning about the content they consume. Any healthcare marketer who can credibly claim that the patients, stories, and experiences they make accessible are true examples of living with a health condition will be certain to get their audience’s attention.
The story of artificial intelligence can be scary, as in the case of deepfakes, when AI-generated content is passed off as a reflection of reality. It can also be inspiring, as in the case of Randy Travis. It’s up to us to use it transparently and ethically, so that the disruption that is inevitably going to happen is one that improves the way people deal with health conditions.
