D. Sahay co-wrote this article with Arun Shastri

This post is the second in a two-part series.  

Post_PMSA_2018_Blog_img-925336-editedPresenting at the 2018 PMSA Annual Conference provided us with the opportunity to think through some common myths about AI. Before our presentation, we spoke with colleagues and clients, and settled on the following five persistent myths:

  1. AI is only relevant for Facebook and Amazon. Stakeholders are familiar with consumer apps that enable image recognition on Facebook or natural language processing with Alexa. When it comes to enterprise business uses, we don’t necessarily have millions of images or much use for them. Instead, we have a lot of structured and unstructured data, and AI can be leveraged in this environment as well.

  2. AI is continuously learning. We frequently hear the assumption that AI is always learning, that the algorithms modify themselves continuously, and if it’s not doing this, it isn’t AI. In reality, very few algorithms function this way. Yes, there’s a component of learning in most AI, but such learning happens on an infrequent basis.

  3. AI is for every problem. A natural tendency after discovering a new tool such as AI is to look in the immediate vicinity for problems that are currently being tackled and ask if it can be solved with the new tool: “How can we leverage AI to improve what we’re already doing?” But that may not always be relevant or the most impactful. Carefully consider the problem and determine if, in fact, AI will enhance the solution.

  1. AI means data scientists. While it’s true that data scientists are essential for creating an AI capability and the algorithms that drive it, we need team members with different skill sets, too. For example, before data scientists create the algorithms for a solution, we need data engineers to gather and wrangle data, and once the algorithms have been developed, we need software engineers to develop and maintain the operating system. If the solution is being presented to the sales force, we may want learning and development professionals to play a critical role in driving adoption.

  2. Data has to be available in plain sight before you can use AI. Netflix didn’t own a massive database full of people’s movie preferences. The company created an infrastructure that gathered the data as users watched videos, and then used that data to predict preferences with AI. And on other occasions, data exists but is hard to find within the organization, and the process to discover it is essential.

Beyond stakeholder management, successfully leveraging AI involves creative problem-solving, disruptive thinking, patience and the right team. It’s not a question of if but when much will be done with AI in commercial life sciences. We’re happy to see so many of our clients embracing artificial intelligence.


BLOG POST: Five Critical Steps for Successfully Leveraging AI

EVENT: PMSA 2018 Annual Conference

BLOG POST: The Keys to Pharma’s AI Success Include ‘Disruptive Collaborations’ and a Willingness to Innovate and Iterate


Topics: AI, artificial intelligence & pharma, PMSA 2018