Artificial intelligence and machine learning (AI/ML) are transformational forces in healthcare. With the explosion of patient data from genomics, electronic health records, wearables and other sources, AI/ML is becoming essential to make sense of this data and answer some of healthcare’s most challenging questions.
While these technological advancements offer greater insights, pharma leaders struggle to implement and scale AI/ML across their enterprises. A Harvard Business Review found that only 8% of chief executives successfully led enterprise-wide AI initiatives, indicating the pace of adoption is slower than the speed of AI/ML advancement. Organisations tend to take similar paths – hire expert data scientists, amass mountains of data, and implement pilots to test small, siloed AI/ML hypotheses. While these are apt first steps, they result in proofs of concepts that stall and return little to no value, despite considerable investment.
Pharma leaders can successfully leverage the value of AI/ML and drive successful strategies across their enterprises while managing the disruption that comes with these capabilities. Here are seven principles that offer a framework for success:
1. Define and win with a molecule-to-market AI/ML strategy
AI is a bridge from clinical to commercial that drives significant time and cost efficiencies. For example, AI can reduce clinical trial time with a 40% faster site identification process and 30% faster recruitment. Other AI proof points include precisely determining burden of disease, managing pipeline investments based on addressable market identification, improving physician targeting and multichannel marketing optimisation and calculating ROI for feedback loop to investment planning.
With more than 50% of molecules in pipelines in specialty areas, finding precisely the right patients is crucial for successful trials and commercialisation. Organisations that implement a coherent AI strategy in every phase – from clinical to commercial – will be the industry winners. They will have the competitive advantage to predict outcomes in one trial phase and execute based on that information in the next phase. AI/ML is not an isolated solution, but rather a capability of the end-to-end journey that compounds benefits throughout the life-cycle.
2. Assign accountable business owners to define business objectives
In not-so-successful cases, AI strategies are assigned solely to technology owners or to several owners with shared responsibilities focused on experimentation. For an AI/ML strategy to be impactful, it needs accountable business owners with tangible goals for business outcomes.
AI has proven to deliver efficiencies via automation, cost reduction and business growth: reducing enterprise costs by 10%; using an AI-led marketing approach to deliver $100 million in revenue for one brand in one market; uncovering addressable market opportunity valued at $500 million for one brand in the US; and implementing AI strategies for commercial operational excellence.
Be clear about how AI can help your business. Assign a leader who can identify AI/ML applications across your organisation, and work with technology owners to drive implementation.
3. Have a plan to scale if you have an appetite to pilot
Several organisations start their AI/ML journeys by adopting a pilot strategy. Leaders list hypotheses and compare the AI/ML application against traditional analytics to prove its value. However, when the minimum viable products are successful, product sponsors are ill-equipped to scale. While this approach is motivated by a desire to minimise investment risk, it is a poor way to drive real value.
Leaders designing pilots need to include plans to scale in the original pilot design. This approach enables leaders to account for necessary choices in methodology, technology, staffing and change management to properly scale and realise value across the enterprise.
4. Use fit-for-purpose data
An AI/ML algorithm doesn’t need all available data to predict outcomes. Organisations often invest in multi-year programmes to collect and cleanse hordes of data to use as inputs for AI/ML models, then discover that model is still not good enough. These data-heavy AI/ML initiatives increase complexity and ultimately fail or are delayed.
While centralised data capture, management and governance are important, adding more data doesn’t necessarily improve predictions. ML algorithms need fit-for-purpose data selected according to its relevancy for each use. Leaders who have led successful AI initiatives have invested in AI platforms and have AutoML and robotic processes to deliver viable fit-for-purpose data for each use. Developing automated mechanisms to clean, transform and integrate data at scale for multiple AI/ML uses drives efficiency and results across enterprise demands.
5. Operationalise your AI/ML models
Pharma has a large concentration of data scientists who build hundreds of models every day. These data scientists develop great models for the same problem, but how do leaders choose the right AI/ML model for a business question?
How many of these models feed into the strategic decision-making, drug discovery or commercial execution of global brand teams?
How do leaders make the model a part of the production process?
Operationalising AI/ML and writing production-level code for data scientists is essential. Pharma leaders should invest in operationalised solutions and set up the end-to-end production processes to enable AI/ML to be embedded in pharma operations.
6. Be prepared to manage the disruption in your organisation
Pharma has traditionally been a conservative industry with limited examples of fast, widespread adoption of innovation. This culture makes embedding AI a change-management challenge.
Leaders should not underestimate the disruption that AI brings, particularly to employees, and should be prepared to manage it. Examples include sales teams that may mistrust the recommendation of an AI algorithm based on their knowledge of the client, opposition from service teams that may feel disrupted by automation and business teams that may need a better understanding of AI before promoting downstream actions.
Leaders should instil an agile mindset where AI products are built iteratively and incrementally, implement educational programmes to help employees upskill for opportunities AI creates, provide incentives for change and organise and budget for cross-functional teams to form around AI initiatives. It is also important to openly address redundancy concerns, reinforce the anticipated value of AI adoption and, finally, progressively instil the openness to use and culture to act on the outcomes of AI solutions.
7. Partner with an expert AI/ML organisation
If you want to go fast, go alone. If you want to go far, go together. Organisations that successfully implement and scale AI/ML initiatives understand the importance of choosing the right partners rather than multiple siloed vendors. A Gartner report found that 48% of CIOs report they have an ‘incorrect vendor strategy’, which costs organisations millions and delays critical initiatives.
Pharma companies and other organisations have brought together technology players, providers, commissioners, patient groups, professional clinical bodies, academic and research institutions and regulators to galvanise more ambitious programmes and achieve large-scale roll-outs of AI and/or the data that enable it. Highlights include:
Collaboration for Oncology Data in Europe, a multi-stakeholder, multidisciplinary, collaborative group with a shared goal to drive innovation and support improvements in cancer care by creating the Oncology Data Network providing near-real time information on how patients are currently being treated with anticancer medicines at scale.
Observational Health Data Science and Informatics, an open-source community that enables the validation of AI/ML through standardised data models across several healthcare data sources and geographies.
Accelerating Therapeutics for Opportunities in Medicine, a consortium transforming drug discovery by accelerating the development of more effective therapies for patients. Some of its central tasks are focused on advancing AI adoption by pharma.
AI has arrived in healthcare, and recent examples demonstrate how it can transform the status quo. Pharma leaders must have the temerity to believe and invest in the power of AI to improve and accelerate the development and commercialisation of new therapies for the right patients. Achieving balance between scientific rigour and business value in AI implementation is important. Pharma leaders should maintain a pragmatic approach to how these capabilities will impact their organisations for the trees of AI to yield fruit.
