Generative Artificial Intelligence (AI) has the potential to be a key driver of competitive advantage within policy, access, value, and evidence; in this update, Onisim summarizes the rise of generative AI, shares his 4-step approach for adopting generative AI, and summarizes a comparison of off-the-shelf tools versus in-house development of curated AI solutions.

 

The rise of generative AI

Generative AI technology creates material such as images, text, audio, and video; and with recent advances in user interfaces, material can be produced faster and with higher quality than ever before.

Leading technology companies such as Google, Microsoft, and Amazon are invested in the field, with each racing to innovate and leverage generative AI for competitive advantage.

At the 2023 Microsoft Build Conference, the tech giant predicted that, in no more than 3 years, anything that is not connected to AI will be considered broken or invisible.¹ Sequoia Capital, the American venture capital firm that specializes in technology growth sectors, suggests that by 2030, Generative AI will be used to write papers better than human authors and create images better than professional designers.²

 

Generative AI in text and images: the future is here and it’s evolving rapidly

 

A 4-step approach for adopting generative AI within HEOR and market access

We can assume that generative AI is here to stay and will have an impact across industries—including biopharmaceutical HEOR and market access. But how should you approach your exploration and adoption of these game-changing technologies?

For HEOR and market access professionals interested in leveraging generative AI, I suggest you think big, start small, and learn quickly by working through these 4-steps; please do contact me if you want to discuss these in relation to your own business goals.

1. Exploring
2. Piloting
3. Scaling
4. Innovating

 

Exploring

To build an understanding of how AI has been successfully implemented, a hands-on understanding of the broad technical landscape of production-ready and near production-ready capabilities and use cases is required.

You could research these yourself or employ an expert consultancy to help you learn quickly about the approaches, frameworks, and design patterns that work and those that don’t.

This phase of technical discovery helps you create an actionable generative AI strategy and prioritized plan with clearly defined business value. It reduces the eventual development time and cost as well as building your understanding of what is possible in the future. It also helps you to de-risk the eventual deployment of new technologies—you will have a greater degree of confidence that you won’t be hitting dead ends or using approaches that will not align successfully with evolving use cases.

 

Piloting

Establishing the value and viability of possible use cases requires experimentation and validation work. You can experiment with “quick and dirty” prototype solutions that tease out gaps in capability and limitations in initial approaches (an incubator-style approach).

For me, the sweet spot for generative AI use cases is where human effort to achieve the desired result is high, while the validation to check the plausibility or correctness of the result is easy.

It is worth remembering that the ability to experiment and pivot rapidly is key to maximizing your ability to validate the viability/value/impact of possible use cases and avoid expensive dead ends.

 

Scaling

As knowledge of and confidence in generative AI grows, you will start to infuse large language models into a variety of business processes and applications.

There are many industry-agnostic use cases such as mining insights from transcriptions, automating document processing, and generating personalized and contextualized content.

 

Innovating

To get ahead of the curve and to differentiate your value propositions with the help of generative AI, you will need HEOR- and market access-specific AI copilots. By combining pre-trained foundational models, such as GPT-4 on Azure, with deep industry and domain expertise, you can develop AI copilots supporting your strategy.

A human-in-the-loop approach is key for successful adoption of innovative AI copilots where a collaborative space is created between the user as a pilot and AI as copilot.

Generative AI technology: off-the-shelf tools versus a curated approach of in-house development

I am speaking with an increasing number of leaders in HEOR and market access about generative AI, and a question I am often asked is “should I invest in off-the-shelf, publicly available tools or develop my own curated approach of in-house technology that is specific to my needs?”

While every case is unique, there are some common points to watch for, such as becoming locked-in to vendors, concerns about data management, and potential for competitive advantage, that I highlight in this figure.

 

Generative AI: public tools versus a curated approach

This, of course, does not mean excluding public tools—some use cases may be well served by third-party solutions. However, attention should be given to potential lock-in and the ability to own the competitive advantage. Prompt engineering is the secret to all large language models interactions, and through a curated approach the opportunity for creating IP is retained by being able to protect value elements.

Additional reading: How could artificial intelligence re-shape market access and launch strategies for new therapies?

As part of this update, I’d like to highlight the research by Avalere Health, which has revealed four focus areas where AI could enhance market access and launch strategies for new therapies: patient stratification, predictive pricing, omnichannel marketing, and targeted promotion.