Back in June, I saw this headline Sanofi goes All-In on AI flash by on Forbes. The author, who does turn out to be the CEO of a Sanofi partner, was pretty bullish on the commitment that Sanofi is making to becoming ‘the first pharma company powered by AI at scale’. I was intrigued enough to dig deeper. As an interested industry outsider, it does seem to me that there are many possible avenues for AI augmentation within the fields of drug discovery, drug production supply chains, etc.
The Forbes article makes a couple of points that I found very interesting.
Point number one - timing in AI investment is everything and while Novartis ran at AI too early (2018) and flamed out, Sanofi, who seem to have started a fairly big push on data and AI about two years ago, is timing the cycle right.
By 2020, it felt like Novartis had more heads of AI than the largest AI drug discovery companies had AI scientists. I remember coming and presenting there several times but it seemed like some of them were constantly on vacation or working on AI strategy, trying to fill the newly-created head counts, or saying that they don’t know who to partner with because they did not understand the landscape of AI startups and they need to run pilots.
Alex Zhavoronkov, Forbes expert contributor, AI for healthcare
If true, this certainly sounds a lot like my experience in both the financial services and telecommunications sectors, where AI has been the darling of the talking classes without a lot of executional capability or grassroots investment in the data infrastructure needed to put a cake under the AI icing as it were.
Point number two - ‘all in’ actually does seem to have some thought behind it with very sensible sounding use cases (albeit many of them smart data plays rather than AI) identified across decision making nudges for internal company functions, drug discovery, vaccine research, clinical operations and manufacturing & supply. Which sounds like a good part of the cost base for a pharma company.
This is the Sanofi CEO, Paul Hudson in the June 2023 ‘all in’ press release
Our ambition is to become the first pharma company powered by artificial intelligence at scale, giving our people tools and technologies that focus on insights and allow them to make better everyday decisions. The use of artificial intelligence and data science already support our teams’ efforts in areas such as accelerating drug discovery, enhanced clinical trial design, and improving manufacturing and supply of medicines and vaccines. We have just scratched the surface as to how we embrace these disruptive technologies to achieve our ambition of transforming the practice of medicine.
Paul Hudson, Sanofi CEO
Then last week, Paul Hudson was in the news again on Yahoo Finance (which is still a thing!), again speaking with a lot of conviction of his belief that AI will be critical to to big pharma from now on.
According to Paul, Sanofi looks at AI in two broad tranches
Expert AI: handcrafted from huge data sets, requiring significant computing power. Worked on by a small number of highly skilled people designing medicines to attach to targets.
Snackable AI: The day to day use for people who want to be nudged into better decision intelligence
He is apparently, a regular user of the snackable AI (which sounds a lot like smart, well built BI) so hats off to him for apparently actually practicing what he preaches. That right there gives me some confidence that Sanofi might actually stay the course to get something useful out of all of this. Executives so often refuse to use the tools directly themselves which rather defeats the purpose and is a sure sign that investment will run out before any real culture change towards data assisted decision making can ever happen.
Both Paul and interviewer Anjalee Khemlani comment that there is a growing need to plan out the entire pipeline development as a process and I think this ties back convincingly to the press release areas of AI application which includes operations, manufacturing and supply.
Nimble AI pure plays in biotech?
At the end of the piece, Anjalee probably nails a big part of this renewed push into AI by big pharma in general when she says
Of course, the company is now competing with all of these smaller startups that are in biotech that is solely focused on the utilization of AI, so it's really on big pharma to really pick up the pace in how frequently they use it, how much they're going to be using it, and what the benefits are.
While not exactly a small startup, that’s exactly what the Google Deepmind team working on AlphaFold is betting on. While this recent article in Nature, casts some doubt on just how useful protein folding predictions are on their own for the drug discovery pipeline, Deepmind, Recursion and Charm Therapeutics are all companies that sit outside traditional big pharma but are squarely focussed on taking a piece of the drug discovery pie. My limited understanding is that they would still partner with more traditional players to bring drugs to market but it will be interesting to see whether power shifts.
But … the talent shortage
When a useful drug exists, of course you have to actually produce it. That’s where the CDMOs or Contract Development and Manufacturing Organizations come in. CDMOs are organizations that serves the pharmaceutical industry and provides clients with comprehensive services from drug development through manufacture. It isn’t difficult to see a myriad of applications for AI here, from computer vision assisted quality control to in-line process improvement via the coupling of spectroscopic monitoring and ML algorithms.
So it was a bit disheartening but also oh so familiar, to read that small and midsized CDMOs are struggling to attract the skilled folk required to enable automation of production lines, visual inspection, quality control and processing, and preventative maintenance of equipment.
“… whatever few PhD data scientists are out there, they are lapped up by the big tech giants, like Google and Microsoft. So, there is not very many data scientists that are available for the manufacturing and operations spaces, which is important if we have to use AI moving forward.”
Amita Quadros, GBI Biomanufacturing
An application of AI that isn’t selling advertising!
So I guess the commitment to ‘all in’ remains to be proven out. (I’m not sure that Google is really living up to it’s intention of being AI first either.) But credit to Paul Hudson, Sanofi CEO for being a visible champion for using the decision assistance tools he is asking others in the company to use.
And while the privacy implications of this make me a little panicky, it’s interesting to see that AI might even be able to help us recruit appropriate patients for drug trials more quickly, shortening the last mile as well as the first.