AI in Biopharma Manufacturing: Why Build-or-Buy Is the Wrong First Question

Yeah.

Very interesting.

Whether it's sponsors, CDMOs, or life science tech companies, there's so much that can be done when it comes to AI and digital.

There's so many problems to solve.

Where do you think it's important to focus and to decide these are the key capabilities which are strategically important enough to buy or to build?

Yeah, I think same as with what we've seen with bioreactors, right? There are some modalities, there are specific workflows, there are places where...

perhaps the expectation is that the market, the rest of the market will not have a need like this.

So a technology company, a specialist will not make something like this and then you have to make it at

home.

But I think most of the time we will land in a place where vastly those needs will be served by specialists or at the very least companies will be building those in collaboration with, with specialists.

It certainly seems whenever I talk to people, there's many, many different pilots going on in all sorts of different areas around KPI dashboarding or batch scheduling, supply chain forecasting, contract maintenance, fragmented tests and pilots, which people are getting up and running.

Of all those different types of pilots that people could be running, where have you... seen the most value from those types of pilots? And also, what does it take to go from a pilot to something which is more scalable and more commercially meaningful?

Yeah, I think it's important to acknowledge that this phase has been a phase of experimentation and companies have taken great effort to explore where the productive use cases are.

And very importantly, to bring the workforce together together with them on this journey of technology transformation.

But

I think it's also clear from our previous experience with other technology transformations as well as with other corporate functions that experimentation needs to move to a more prioritized and structured portfolio level view of these initiatives.

So going back to our earlier discussion about making a decision to build or buy or build, I think that goes back to what we're talking about now, where maybe in this exploration stage, there's something that you deploy as a pilot in house, but That's only to demonstrate that there is value in this pilot.

And then you actually bring in a specialist to do this at scale with validation in mind, compliance, security, and so on.

Starting small to confirm that there is a business need for it.

Yeah, or that it's even feasible, that the amount of effort, resources, time that it will take to prepare the data, standardize the data, make it AI ready is actually a feasible exercise.

I think that experimentation phase has pointed to a number of

gaps.

let's say in the data itself.

But I think at that portfolio level, what needs to come in as a next step is well-structured prioritization, resourcing, and most importantly, trade-offs for the projects that will and will not move forward.

So I think there from the customer side, what we're seeing is that there's evaluation of these AI pilots sort of one at a time, They get approval, they move forward, which is again fine for the experimentation stage.