THE HIDDEN RISKS AND COSTS OF MANUAL AI IN PHARMA
The delivery of AI insights within Pharma organizations remains a manual endeavor characterized by:
Constant one-off data integration
Data scientist-driven execution
Limited quality assurance
Given that standard data and reporting processes are automated (operationalized), why is AI being treated differently? There are some strong influencing factors. First, the answers generated have a unique and positive impact on the business. Second, AI is far more complex than standard reporting. And finally, the effects of manual AI do not directly impact the financial statements.
These are all reasons for executives and analytic leaders to hold off on changing the status quo but what is manually executed AI really costing organizations? What is it exposing them to? This article outlines the hidden risks and costs of manual AI and how these apply not only to internal efforts, but also to the vendors that provide AI.
COMPLIANCE RISK - In Pharma there are many regulations. Data privacy regulations, messaging regulations, manufacturing regulations, and label regulations (to name a few). And there will soon be AI regulations. No human can keep track of all the dimensions to compliance. This is why you are seeing compliance pop up as a prime use case for GenAI. The only way to effectively manage regulatory risk is to harness the power of algorithms and computers to sift through all the potential gotchas.
HUMAN ERROR RISK - Data scientists exist to build models that produce the best possible answers and they do the best job they can on QA. There are a host of potential sources for errors in AI results from bad data to incorrect machine learning settings made at run time. Many times the errors are only identifiable by business experts. The data scientist cannot be expected to effectively perform all the necessary checks every time they run an analytic and errors frequently make their way into the results.
DEPENDENCY ON INDIVIDUAL EMPLOYEES - This is the simple but scary effect of manual AI. If the data scientist is the only person who knows how to run the code, and they leave, what happens to the business processes and decisions dependent on that AI?
LIMITED BUSINESS SUPPORT - When people think of automation they associate it with speed. An automated process runs much faster than a manual one. One of the major hidden costs of manual AI is that answers cannot be generated fast enough to support constant and rapid decision making.
HIGHER TOTAL AI DELIVERY COSTS - Let’s start with the proposition that you have an AI based analytic that is being run period over period for a brand. Next best action (NBA) is a good example. Now you want to run NBA for several more brands. If NBA results are being generated manually for the original brand, how do you scale? Do you hire more data scientists? Do you end up with wildly different NBA analytics for each brand? Independent methodologies across brands and the siloed data scientists necessary to run each are far more expensive than a unified and operationalized approach.
REDUCED INNOVATION - Lastly, if data scientists are running AI and performing all the checks for compliance and errors, who is developing the AI that will drive new business opportunities? By overloading data scientists with operational tasks, organizations are reducing the innovation that comes from such a valuable resource.
The challenges associated with manual AI in Pharma tend to be overlooked without a ready means to address them. Even more detrimental: AI models that can produce profound patient outcomes and business results are not utilized because of the potential risks.
One common approach to sidestep the internal challenges associated with AI is to have external vendors provide the answers. But ask yourself this: why are any of the challenges above different for a vendor that supplies AI to your organization? If you are using a vendor that does not automate their AI processes, you are exposing yourself to the same risks and costs? A model can only be considered complete (internally or externally) once it has been operationalized.
----
We welcome the opportunity to tell you about our industry leading AI driven analytics for Pharma and the results we have produced for our customers as well as how VELOCITY, our analytic operations platform, mitigates the risks and shortcomings of manual AI for our models and can do the same for yours. Please schedule an introductory meeting here.
