There is no doubt Artificial Intelligence (AI) is having tangible benefits across the entire pharmaceutical chain. But there is also often a lack of clarity around AI. This can lead to an overestimation of potential impacts and underestimation of the complexities of implementation.
In this article we compare the hype surrounding AI with the reality and set out the practical considerations for successful AI implementation.
What is AI?
AI is the ability of machines to perform tasks which have typically required human intelligence. These include learning, problem solving and decision making. An advantage of AI which makes it particularly suited to the pharmaceutical industry is its ability to process large amounts of data and identify patterns.
Machine learning (ML) is a subset of AI focused on developing algorithms and statistical models. These algorithms enable computers to learn from data, make predictions or decisions, and improve as they receive more quality data without being explicitly programmed.
Recently, advanced ML models called large language models (LLMs) have been used in generative AI (GenAI) to create text. LLMs can process and generate more human-like language than previous AI methods and public awareness has rapidly increased due to accessible models like Chat GPT.
However, the buzz around AI has created its own issues – from unrealistic expectations to concerns over job losses. So, what is the hype and what is the reality?
Hype Versus Reality
AI is sometimes presented as a silver bullet – the solution to all our problems. While there are undoubtedly exciting opportunities on offer, the adoption and integration of AI with existing datasets will not be seamless. Successful adoption requires more than just choosing the best AI model or solution. It requires careful change management, including addressing legitimate concerns, training, and stakeholder engagement. Effective data architecture and technology infrastructure, along with close collaboration between business leaders, legal and risk teams will also be crucial to allow important insights to be unlocked.
Amidst concerns about monopolization and data security, initiatives such as the FAIR Principles (Findable, Accessible, Interoperable, and Reusable) will be key in ensuring data is easy to discover, share, and integrate over domains.1
There have been widely reported concerns that AI will lead to job displacement. However, the reality is more complex. A report from the IPPR think tank says the impact of AI adoption on the UK labor market will depend on policy.2 The worst-case scenario is full displacement with all at-risk jobs replaced by AI, 7.9 million job losses and no GDP gains. The central scenario is 4.4 million jobs disappear, but with economic gains of 6.3 per cent of GDP. The best-case scenario is all atrisk jobs being augmented to adapt to AI, instead of replaced, leading to no job losses and an economic boost of 13 per cent to GDP.
The truth is AI is only as good as the people who run it and the “human in the loop” is vital for successful implementation, execution, interpretation, and decision-making. AI, like any technology, has limitations and risks. Skilled professionals and subject matter experts (SMEs) are needed to identify and overcome errors, bias and data quality issues that could affect performance.
How is the Industry Already Using AI?
Top pharmaceutical companies are already using AI to gain a better understanding of diseases, predict future patient responses to treatment, pioneer new clinical approaches and accelerate analysis.3-5 A review of manufacturing sites found facilities using a range of digital technology and advanced analytics saw a 25-40% increase in plant capacity and a 15-20% reduction in lead times.6
In Phase I clinical trials, AI-discovered molecules have an 80-90% success rate, substantially higher than historic industry averages.7 AI systems have also been trained to score tumor cells and immune cells for a biomarker, called PD-L1 – a development with the potential to help inform immunotherapy-based treatment decisions for bladder cancer.8
AI is even revolutionizing precision medicine with biomarker testing helping doctors and patients identify the treatment plans most likely to work for them9 and significant applications for big data and AI in oncology.10
Considerations for Successful AI Implementation
So, how can we continue to effectively integrate AI and unlock new opportunities? The answer lies in the interplay between both technical and organizational aspects, supported by a clear strategic plan.
For organizations with no or limited in-house technical expertise, selecting an experienced partner is the first step in integrating and unlocking AI’s capabilities. The correct partnership can help the organization to select the right technology to meet their specific needs. Collaboration helps create a clear problem statement, long-term roadmap, and a precise set of requirements, leading to the effective identification and evaluation of available AI solutions. An effective collaborative partner should consider the costs and ease of integration with any existing systems. If no off-the-shelf solution meets the requirements, the chosen expert partners can help to create bespoke solutions and deliver value.
Good data is crucial for successful AI implementation. Organizations should have effective processes in place to ensure data quality, completeness, lack of bias and accessibility. Policies should also include rigorous and ongoing validation to establish and maintain model trustworthiness. There should also be a robust risk framework and effective governance to ensure ethical AI practices.
As we have already discussed, successful AI implementation is about much more than just choosing the right model. It also requires business transformation and effective change management, which in turn takes time and careful strategic planning. Business leaders must be aligned on objectives and how those objectives help meet the organization’s overarching goals. There should be clear success metrics which are regularly monitored to understand the system’s impact and value. Clear feedback mechanisms and a culture of ongoing improvement are also vital to ensure systems remain fit for purpose.
Early and ongoing engagement with stakeholders is crucial to address concerns and support successful implementation. This should include effective training to help employees understand new systems and, where necessary, job augmentation. General AI training can also be useful to create a responsible use of AI culture within the organisation and upskill existing employees.
Conclusion
To unlock the benefits of AI, companies must approach its adoption with a critical mindset, differentiating the hype from tangible benefits and genuine opportunities. While aiming for a return on investment (ROI) leaders must be open to embracing new innovations and be prepared to adapt and evolve their strategic roadmaps as the technological landscape progresses.
We must appreciate that AI is only part of the solution – using the skills of SMEs, goal-driven strategic planning and collaboration will also be key.
By taking these steps, we will be able to harness the strengths of AI and accelerate access to transformative treatments for all patients.
References
- https://www.go-fair.org/fair-principles/
- https://www.ippr.org/media-office/up-to-8-million-uk-jobs-at-risk-from-ai-unless-government-acts-finds-ippr
- https://www.astrazeneca.com/r-d/data-science-and-ai.html
- https://www.gsk.com/en-gb/behind-the-science-magazine/ai-ml-data-computing-power/
- https://www.novartis.com/stories/ai-changing-face-healthcare
- https://www.mckinsey.com/industries/life-sciences/our-insights/reimagining-the-future-ofbiopharma-manufacturing
- https://www.researchgate.net/publication/380223979_How_successful_are_AI-discovered_drugs_in_clinical_trials_A_first_analysis_and_emerging_lessons
- https://www.astrazeneca.com/r-d/data-science-and-ai.html#UsingAI
- https://www.gsk.com/en-gb/behind-the-science-magazine/biomarker-testing-ai-tumourshepatitis-respiratory-cancer/
- https://www.astrazeneca.com/what-science-can-do/topics/data-science-ai/unlockingpotential-data-ai-driven-drug-discovery-development.html#
Author Details
Dr. Jennifer Bradford, VP, Head of Data Science, Coronado Research
Dr. Jennifer Bradford has a diverse background spanning academia, Contract Research Organisations (CRO) and Pharma. She has a proven track record in leading the development and implementation of customer-centric digital solutions, including AI and data science, within the industry. With over 20 years of experience at the intersection of digital innovation and clinical research in both the pharmaceutical industry and academia, she has been instrumental in leveraging advanced digital technologies to deliver value to organisations. Her career began in proteomics research and has since evolved to include clinical drug development and pharmaceutical applications, where she has consistently bridged the gap between scientific knowledge and the practical implementation of technical solutions
Publication Details
This article appeared in Pharmaceutical Outsourcing:Vol. 26, No.2 Apr/May/June 2025Pages: 8-9