ChatGPT and generative AI land in pharma marketing with sense of awe — and plenty of caution

A professionally dressed woman stands in front of an office background and introduces herself to talk about the use of artificial intelligence in pharma marketing.

Except she’s not a real person. “She” is an AI-generated avatar in a video created in less than a minute using the generative AI application Synthesia along with a typed script from a real person — in this case, a healthcare agency executive showing an example of how a generative AI avatar can work.

Generative AI — the broad category for recent headline-generating applications that include ChatGPT, DALL-E and Synthesia — can create original articles, social media content, illustrations or instructional videos in minutes. And pharma marketers and healthcare agencies are experimenting and beginning to use it just like consumer companies, although with an added level of caution.

Simon Wilson

“It’s brilliant, it’s world-changing — and it’s free at the moment, which is why a lot of people are using it — but it’s also not infallible,” said PrecisionEffect innovation director Simon Wilson, who created the earlier mentioned AI avatar video.

That’s in part because generative AI relies on data sets that gather human-created content, which can be inaccurate or biased. Large language models like ChatGPT can also omit details and even “hallucinate,” as Wilson explained. That means in attempting to bridge gaps, the applications may sometimes try to be creative and “make up nonsense,” he said.

Eversana Intouch CEO Faruk Capan agreed about the need for caution and said, “What pharma companies and companies like us need to do to make this work is to ‘pharmatize’ it. Make sure it’s failproof, foolproof and factually correct and looked at by [real] people. But it’s a good starting point right now.”

Faruk Capan

That’s why generative AI, for now at least, is selectively and carefully used in pharma marketing. However, it is definitely being used. Most often, behind the scenes to not only aid and abet creativity but also simplify production and cut costs.

Faruk Capan AI avatar

Capan offered an example Eversana Intouch is testing with generative AI used to create expert videos for continuing medical education (CME). Instead of flying around the world to videotape its advisory board health experts, Eversana can shoot a few seconds or minutes of video of each person as a “talking head” and then use generative AI to create longer educational videos. The agency feeds the copy — approved by the physician or expert every time the script is changed, he noted — into the generator, which then simulates the expert’s voice, syncs mouth movements to the words and creates an original video.

Capan said, “It’s going to impact everything we do,” in an echo of a sense of awe among pharma marketers and the broader population around generative AI right now.

While seemingly hyperbolic, the sentiments also come with an insider understanding of where generative AI fits — mainly as production assistants for now. Capan himself is working on an internal presentation with the working title: “AI won’t replace you, a person using AI will.”

And that’s at the core of generative AI, especially for pharma marketing. The current whizbang software applications can process enormous amounts of data, spot patterns with unmet speed and respond with original human language scripts or authentic aesthetic images. However, it still doesn’t always get it right. And that requires human oversight for accuracy and empathy.

For example, a text request to ChatGPT to write a TV commercial or jingle for a new diabetes medicine returned several scripts that included words such as “cure” or “revolutionary.” Any pharma marketer would know that promises like those could get a brand in trouble with federal regulators.

Emerging medical-specific generative AI applications, such as Google and DeepMind’s recently announced large language model MedPaLM, along with custom applications created for a specific pharma product, brand or health condition, may help with both factual errors and regulatory knowledge.

Still today, copywriters and content creators are using generative AI for background tasks like brainstorming, examples of what not to do, and even competitive and analytic data sets.

Alfred Whitehead

“These AIs are extremely good at reformatting ideas in different styles and new ways of thinking about them, which we found is a good accelerant for brainstorming. It gets people more engaged and people can ‘see’ what they’re talking about right away,” said Alfred Whitehead, Klick EVP, applied sciences.

He added: “It reminds me of the early days of the web where all of a sudden we had something that was clearly new and clearly powerful — and very exciting. Everyone’s trying to figure it out and what’s the right way to use it.”

One issue still under discussion is how transparent marketers will need to be. With much of the content generation still behind the scenes and not going direct to patients or physicians, the question of how, or even if, to disclose the fact that the content or a video is AI generated is still being debated.

CMI Media Group’s Mark Pappas, SVP, innovation, and Franco Maffei, VP, search engine optimization, pointed out several ways generative AI can be used.

Pappas pointed out a doctor, a urologist named David Canes, who explained in a Twitter post how he used ChatGPT to write a letter to an insurance company rebutting a denial for a prostate MRI for one of his patients. Canes wrote, “So convenient and efficient. Even gave references!” Although that also means insurers could flip the script and use the technology to explain denials of coverage or drugs.

The CMI execs pointed out other uses for pharma marketers right now, such as the ability to create an accurate profile of a competing brand’s social media presence by listing accounts, channels and content in just minutes. Other generative AI can record Zoom meetings, then transcribe and analyze the content for key themes and outlines or can scan creative content for alternative, improved words, sentences or paragraphs, they said.

Maffei had one of his team members ask ChatGPT how artificial intelligence is going to impact search engine optimization — in preparation for research the agency was doing on the topic.

“He created an entire write-up with key points and main ideas, and it happened within a minute. It was insane,” Maffei said, and mostly accurate.

The potential is almost unlimited, as Pappas noted, “It can do whatever you want. It’s almost terrifying what it can do — but in a good way.”

artificial intelligence
ai
ai generated
software
healthcare
medicine
medical