About three months ago, I decided to take a deeper dive into Artificial Intelligence (AI).
While AI has been overhyped in the media, it does represent a game-changing technology that will make its way into every industry.
As such, how will AI change the game in GMP Quality and Inspection Management?
Before we go there, an important bit of context to remember: new technologies are almost always first applied to the industries and verticals with the greatest financial incentives.
So, it’s no surprise that people are now using AI to transform transportation (autonomous vehicles is estimated to be a $128 billion dollar industry).
In our healthcare neck of the woods, there are at least 16 big pharma (and many start-ups) already using AI to speed up the drug discovery process. This also makes sense, as the drug discovery market is valued in the ballpark of $50-70B.
For an individual company, discovering a new blockbuster drug can add billions to their market cap. What’s the value of avoiding manufacturing quality problems? We’ve written previously about the 5 “costs” of a manufacturing quality problem, some of which can go into the hundreds of millions of dollars range.
We’re talking about 3 orders of magnitude here — use AI to:
|Discover a blockbuster drug
|$1-10B in New Revenues
|Avoid a massive manufacturing problem
|$100M – 1B in Cost Avoidance and Lost Revenues
|Avoid a contained manufacturing problem
|$10-100M in Cost Avoidance
So, logically, AI will start at the top and then eventually find its way down. So prepare for a day where AI will find its way to our world of GMP quality and inspection management. For example, couldn’t we utilize AI to:
- Harness all of this inspection data to predict the next inspection?
- Reverse engineer what the FDA wants to emphasize in the next 12 months?
- Predict which 483s have a higher probability of turning into a warning letter?
- Analyze the real-time front-line manufacturing data and flag variances more quickly?
- Design manufacturing plants and processes more quickly?
- Maximize manufacturing yields with the exact optimal vessel sizes, temperatures, mixing rates, etc?
- “Read” all of our 10,000 483s – what conclusions would it draw?
Okay, it’s fun to speculate about the future. Back to today’s reality.
There is one simple, concrete task that we can begin right now. Investing in this will pay off in the future.
It still may be years before AI hits our industry, but in the meantime, there’s work to do. Why not begin to work on the processes and systems that enable you to get good, clean, consistent data? That data is the foundation to anything AI and can be crunched in new ways for new insights.
It’s this data that ends up feeding the AI neural net (or a good data scientist even without AI!) to identify patterns and insights. But these insights will only be as clear as the data is clean. How can you ensure that your quality organization’s data is collected, scrubbed, and stored effectively?
We are also beginning to explore how we can incorporate AI and data science into our offerings to you.
Our first step is always to collect as much data as possible. We then deploy proprietary processes to clean, categorize, organize, and link the data in as many useful ways as possible.
Now with 20,000+ warning letters, 11,000+ inspection documents, and 700,000 inspection data points, we can already draw actionable intelligence from the data. But this is not the end…just the beginning.
Subscribe to Redica Insights
Get quality and compliance insights from our experts in your inbox