AI Gets Real: Practical AI Solutions for Pharma

It's probably premature to talk about “legacy” artificial intelligence (AI) considering that most pharmaceutical companies have yet to adopt wide-scale applications of AI across their organizations. However, with the advent of OpenAI’s ChaptGPT in November 2022 and GPT-4 in March 2023, we now have a distinctive new category of AI, known as generative AI, that is taking the world by storm.¹ It makes it necessary to distinguish traditional applications of AI from the novel iterations of generative AI, which are proliferating as we speak. Further, it begs the question: What will generative AI solutions for pharma look like?

A Legislative Response to Foster Responsible AI

These new developments are prompting thorough and thoughtful guidance surrounding AI and its uses. The European Union (EU) is leading out, becoming the first legislative body to wrangle with the legal and ethical complexities of AI.² As early as April 2021, the EU proposed a regulatory framework for the AI Act and acknowledged that AI could offer many unprecedented benefits, including better health care. According to European Parliament News, on June 14 of this year, members adopted a negotiating position on the AI Act with the goal of reaching agreement on the final form of the law within the European Council by the end of 2023. As of this writing, the Council has completed final negotiations for rules on transparency, bias, risk, copyright, privacy, security, complaints, fines, and more.³ Once the language is finalized and ratified by parliament, the AI Act will be the world’s first comprehensive set of rules on AI.

Practical Applications of AI Software for the Pharma Industry

One foundational study on AI in the pharmaceutical industry found that small and large companies tend to use AI differently, but that AI solutions for pharma significantly increase efficiencies, and the potential applications are promising.⁴ What will differentiate pharma manufacturers is how robustly they use modern tools like AI-enabled software throughout their processes, including in production and quality control.⁵ Implementing novel technology will need to unfold strategically. The key is to start with low-risk, high-priority business cases, such as managing documents or predicting training lapses, for example, and gradually move on from there.

In a recent report, Deloitte outlines some practical applications and challenges of AI in the life sciences. It states that AI can ultimately serve to “hyper-personalize customer engagement across a multitude of markets and channels,” and offer a “level of granularity and speed that was previously unachievable.”⁶

Here are some practical applications from that report that may be especially pertinent to pharma quality and manufacturing professionals.

• Mitigating financial loss by facilitating compliance with consistency across geographies and topics.
• Reach across data sets to manage supply chain variables with precision and forecast demand.
• Simplifying data access and analysis for business users to accelerate time-to-insight.
• Continually refining domain-specific process models with new training data relevant to user’s specific professional needs.
• Overcoming talent gaps with automation.
• Empowering domain experts without requiring technical proficiency.
• Automating submissions, claims, and appeals.
• Timely complaint and message management.
• Personal assistance for physicians and patients.

Distinctions Between Traditional AI and Generative AI

It is useful to try and break down some of the distinctions between generative AI and traditional AI, or original forms of AI. Both types can utilize now-familiar models, such as machine learning (ML) or large language models (LLMs), but the outputs are unique.

According to Forbes AI contributor Bernard Marr, while traditional, also called Narrow or Weak AI, is designed to respond to a particular set of inputs and perform a specific task intelligently, generative AI “can be thought of as the next generation of artificial intelligence.” He goes on to say that it can be used to “create something new.”⁷

Outlined below are some key differences between the capabilities and applications of these two different categories of AI.

Traditional AI

• Most common AI applications currently powering industry.
• Analyzes existing data.
• Recognizes patterns.
• Used for task-specific applications.
• Trained to follow specific sets of rules.
• Programmed with predefined strategies.
• Powers chatbots and recommendations.
• Performs predictive analytics.
• Supports automation.

Generative AI

• Creates new data.
• Creates original, creative content (text, images, music, codes, etc.).
• Creates patterns and prototypes.
• Reduces time for ideation.
• Supports creation and innovation.

Making the Business Case for AI

According to Marr, OpenAI is drawing substantial investment and long-term partnerships to accelerate AI innovations worldwide.⁸ And OpenAI is just one purveyor. Another AI service provider reports that that nearly “50% of global healthcare companies will implement artificial intelligence strategies by 2025.”⁹ Deloitte experts believe it is crucial for how businesses operate down the line and that generative AI “will likely sit at the core of consumer businesses.”¹⁰ Overall, AI is good news for pharma as it takes on average more than 10 years and billions of dollars to develop a new drug.¹¹ AI is giving pharmaceutical companies “machine vision” into a future reality of faster and cheaper drug discovery, trials targeted to appropriate patients and populations, and the manufacture and distribution of personalized products at scale.

Ethics, Governance, and Monitoring of AI in the Life Sciences

When weighing the current challenges against the future benefits of applying AI in the life sciences, The Deloitte dossier repeatedly returns to four pillars. These pillars will ultimately determine how practical AI solutions for pharma or any other industry actually become. They are:

• Transparency. Data outputs and processes must be clear, interpretable, explainable, and attributable.
• Accuracy. Data inputs need to be accurate and up to date, and outputs should be validated and monitored.
• Trust. Applications must be safe, secure, respectful of privacy, fair, and impartial. Transparency and accuracy increase trust.
• Adoption. Adoption increases with usability, reliability, and trustworthiness.

Human validation will remain a key component to the responsible implementation of AI solutions for pharma within strict regulatory compliance and legal statutes. Choosing to adopt AI software designed for the pharma industry by a trusted vendor is the first half of the journey. The “iterative enhancement” model inherent to AI-assisted technology should allow business users to complete the second half safely aligned with core ethics, organizational values, and pertinent laws. Over time, any viable model should increase accuracy, utility, and relevance to the user’s specific professional needs.¹² For more information about AI solutions for pharma and other pharmaceutical trends, download the trend brief “2024 Pharmaceutical Industry Trends.”

References:

1. "The Difference Between Generative AI And Traditional AI: An Easy Explanation For Anyone," Bernard Marr, Forbes, July 24, 2023.
2. "EU AI Act: first regulation on artificial intelligence," European Parliament News, updated June 14, 2023.
3. "Five things you need to know about the EU’s new AI Actt," Melissa Heikkiläarchive, MIT Technology Review, December 11, 2023.
4. "The role of artificial intelligence in business transformation: A case of pharmaceutical companies,” Ignat Kulkov, Technology in Society, Volume 66, August 2021.
5. "Tech Trends 2023,” Mike Bechtel and Bill Brigs, Deloitte Insights, 2023.
6. "The Generative AI Dossier: A selection of high-impact use cases across six major industries," Deloitte AI Institute, September 12, 2023.
7. Supra note 1.
8. "A Short History of ChatGPT: How We Got To Where We Are Today,” Bernard Marr, Forbes, May 19, 2023.
9. "AI in the Pharma Industry: Current Uses, Best Cases, Digital Future,” by Samantha McGrail, Pharma News Intelligence, xtelligent Healthcare Media, accessed December 11, 2023.
10. Supra note 6.
11. "AI is dreaming up drugs that no one has ever seen. Now we’ve got to see if they work,” Will Douglas, MIT Technology Review, February 15, 2023.
12. Supra note 6.