Pragmatic AI in Pharma

Ron Wince

It is nearly impossible to miss the hype about artificial intelligence and its potential to transform business performance. But despite the optimism about its potential, artificial intelligence adoption in healthcare and pharma lags nearly every other industry. According to the HIMSS Analytics 2017 Essentials Brief, less than 5% of healthcare organizations are currently using AI and nearly half of healthcare organizations see broader adoption in 2–5 years.

What’s holding back AI in healthcare?  

In spite of estimates from Accenture and McKinsey that AI has the potential to deliver billions of dollars in annual impact, most organizations are still struggling to get the ball rolling. So, what are the barriers to adoption and deployment of AI in healthcare. There are three key factors that are getting in the way:

• For most healthcare organizations artificial intelligence is still “a black box”. Machine learning – a subset of artificial intelligence used in business for decades, is still a relatively new and mysterious set of technologies.  As a result of the unfamiliarity, organizations struggle to pin point the right opportunities for applying the AI technology, quantify the potential impact and  find the right channel and application to embrace the new technology. 

• Much of the current health IT infrastructure and applications were not designed and developed with AI in mind.  Legacy systems and even newer EMRs are purpose built, have siloed data storage and handling architectures, and lack interoperability. Successfully deploying AI requires the courage to integrate data and share – neither of which are currently in the comfort zone today’s health and pharma IT managers.

• Data is the fuel that makes AI work. But the vast majority of healthcare data – approaching as much as 80% – is in free text in medical systems of record. To take advantage of large scale adoption of AI, organizations needs to start with a tactical and strategic plan on how to unlock and curate their data for hosting, distribution and consumption by the technology.

Enter Pragmatic AI

Despite these challenges, every enterprise has a sensible starting point to begin leveraging artificial intelligence. For many, it may make sense to start with baby steps. By taking a highly focused “pain point”-centered approach, organizations can start to see positive outcomes quickly at a lower cost and with minimal risk. Pragmatic AI allows organizations to start with the basics, create immediate business value and ROI…then increase adoption as their comfort, capabilities and confidence with AI increases.

Pragmatic AI can be thought of as a continuum with typically three phases or “Generations” for enterprises to follow:

Pragmatic AI Gen I

Gen I starts with a focus on solving a specific, discrete business problem by leveraging AI to automate core business or clinical workflow processes.  Intelligence automation allows companies to take advantage of lower level, basic AI tools that are easily adaptable and proven to make a tangible impact. Even more, AI moves from being an opaque concept to a transparent and purposeful technology solution with enterprise-wide value and high utility for the business. In addition to capturing real efficiencies by automating manual processes, even complex ones, Intelligent Automation also manages and visualizes data from the applications, internal and external systems, and data sources typically unintegrated in transactional and informational processes.  Automating these processes allows companies to begin curating live data sets generated and makes them available for machine learning. 

Pragmatic AI Gen II

In Generation II, companies continue to expand their use of Intelligent Automation in larger and more complex business and clinical workflows – even automating processes that cross internal organizational boundaries.  But now that the AI has started to lose its mystery, many organizations progress to leveraging more advanced machine learning models and tools such as broader uses of natural language processing, image recognition, automated machine learning, etc. In Gen II, organizations begin to quickly identify use cases and quantify business impact based on the experience gained in Gen I. Additionally, organizations now become accustomed to the benefits of accessing data across the enterprise and begin to identify ancillary technology such as blockchain or decentralized databases which allow for intra-company sharing of data.

Pragmatic AI Gen III

Gen III focuses on extending the use of AI outside of the organizational boundaries to include business partners along the value chain, customers, patients across accountable care networks. While Gen II was focused on unlocking data within the organization, Gen III is focused on unlocking data outside the organization and leveraging both for machine learning and AI. Intelligent Automation can now be applied on cross-entity business processes such as claims, compliance, risk management and regulatory reporting. As a result of their work in Gen I and Gen II – organizations now recognize that no single entity has all of the data that is needed for machine learning – especially to succeed in at-risk reimbursement models, drug discovery or in population health management.   

The hype around AI is hard to miss and the mystery is overstated. Despite the deluge of stories about how machine learning will alter the competitive landscape and shift the balance of power, adoption of the technology remains at an embryo state.   But there is risk in waiting to get started – delaying could mean the difference between staying relevant or being left behind.  Fortunately, Pragmatic AI provides a palybook for organizations to embrace AI at their own pace with proven tools, technologies and immediate, visible ROIs.

Ron Wince is the Founder/CEO of Myndshft, a company working at the intersection of artificial intelligence, automation, and blockchain.