Implementing generative AI in pharma: Safeguard communications and boost efficiency

Organizations across various (most) industries are currently in a frenzy over generative AI, and the pressures to adopt it are great, even in a highly-regulated industry like pharmaceuticals. It’s not hard to see why. Who wouldn’t want the ability to incorporate human-like interactions in natural language into applications, generate well-written content on the fly, and automate the precise personalization of content for specific individuals?

However, as with any cutting-edge technology, there are downsides. Generative AI is well known for “hallucinating” information and references that are completely wrong or don’t exist. And while the rate of hallucinations can be improved, it’s unlikely that they will ever be completely eliminated—it’s inherent in the generative process of creation. In pharma, that’s a big liability because providing doctors and patients with incorrect information could cause serious harm—even death. 

Additionally, pharmaceutical companies face unique challenges and can only communicate with patients within a strict regulatory framework that governs not just the information they provide but also the format and even the font size. AI could likely be trained to work within these guidelines, but it only takes one mistake by AI to land a pharmaceutical company in hot legal water. 

That said, while generative AI is not suitable for use cases in which it would interface directly with users (patients, providers, payers, etc.)—the risks are too great—that doesn’t mean there are no applications for generative AI within the pharmaceutical industry, so long as they are implemented and governed with care. Here are three ways pharmaceutical companies can utilize AI to create efficiency, but given the immense power and flexibility of generative AI, these only scratch the surface.

Document processing and summarization

Pharmaceutical companies must follow many precise and complicated rules and steps in all aspects of the business, from conducting clinical trials to manufacturing, distribution, sales, and marketing. The knowledge and instructions for following these steps exist within the company, but finding and digesting them can be difficult. Generative AI could be trained on this information and then provide answers internally, which could save a great deal of time. 

Generative AI could also be a huge time saver when summarizing research. Research documentation can run hundreds of pages long, and the text is usually very dense. Generative AI can parse through content to identify trends, themes, and major findings. The results would need to be checked to avoid any bias, but AI can give users a huge head start on understanding the critical findings.

Personalization

Generative AI is starting to see rapid adoption in marketing operations across almost every industry because, when paired with other forms of AI and analytics, it can automate the creation of highly personalized communications for individual customers. For instance, if you previously bought flowers for your partner on your anniversary, generative AI could use your purchase history to write a personalized email with specific flower recommendations and send it exactly when you are most likely to purchase them. 

Pharmaceutical companies will likely not be using generative AI for automated customer communications—without the support of humans—any time soon because it’s not safe for patients or themselves. But on an ad hoc basis, generative AI can add a lot of value to communications.

For instance, generative AI could serve as an interface with the customer relationship management (CRM) system so pharmaceutical reps could instantly access descriptions in natural language of recent interactions with specific customers. It could also work in the other direction, with AI automatically taking notes on calls and other interactions, which it would then enter into the CRM system. And on an ad-hoc basis, AI could create personalized communications, though it’s absolutely necessary that reps check their content before sending—full automation should be prohibited. Given generative AI’s predilection towards hallucination, there are big significant compliance and regulatory concerns. Plus, if the AI has access to information in the CRM, there’s always the possibility it could disclose private data to unauthorized individuals. 

Process automation for healthcare professionals (HCPs)

Healthcare professionals’ days are packed from start to finish, and a common complaint among them is the extent to which administrative tasks eat into precious time with patients. Generative AI can help pharmaceutical companies be a good partner to HCPs by automating tasks such as patient enrollment, prior authorization, and even answering frequently asked questions (FAQs)—though it all needs human supervision, especially the FAQ responses, each of which a qualified person should review before hitting send.

Guidelines to follow whenever implementing generative AI

Even in these use cases, which avoid direct exposure to customers and HCPs, pharmaceutical companies need to follow guidelines to keep their application of generative AI ethical, safe, and current. Here are a few of the most critical. 

Monitor AI-generated results closely

Whatever generative AI use cases a pharmaceutical company decides to implement, they must regularly monitor AI-generated content for quality, legality, and compliance. It’s important not to become complacent. Just because all the content for the first few weeks passed muster, that doesn’t mean the AI won’t produce noncompliant content or even outright nonsense in the future. 

Continuously improve the AI model

It’s also important to update and improve the AI model regularly to respond to user feedback, changing requirements, and, especially, advancements in AI. Technology never stands still, and that goes double for generative AI. 

Dario Amodei, who led the OpenAI team that created GPT-2 and later co-founded the AI firm Anthropic, likes to talk about scaling laws, which project how generative AI improves exponentially as it trains on more data. The upshot is that AI capabilities are advancing much faster than most people realize. As Ezra Klein said in the preface to his April 2024 interview with Amodei, “Amodei believes we’re just getting started, that we’re just hitting the steep part of the curve now. He thinks the kinds of systems we’ve imagined in sci-fi, they’re coming not in 20 or 40 years, not in 10 or 15 years, they’re coming in two to five years.”

Have a plan to mitigate AI crises

Consider and develop strategies for how your AI will handle edge cases or unexpected input. It’s crucial to put guardrails in place so the AI doesn’t just start riffing when it doesn’t have enough information to provide a grounded answer to an edge case or inappropriate question and so it knows how to respond appropriately to inappropriate prompts. And, should the worst occur and AI-generated content goes awry, you’ll want a crisis management plan already in place to rectify the situation. 

Implementing generative AI in pharma requires thought, not automation

Generative AI is powerful, but it should not be used in an automated way for patient- or HCP-facing use cases. However, generative AI can still be used internally—with supervision, planning, and guardrails—to improve efficiency and accuracy.

At Think Company, we believe digital solutions should be people-powered and AI-enabled. Is your pharma organization utilizing generative AI effectively and appropriately? Let’s chat.