I spent late April in Barcelona at Reuters Pharma Europe, and one sentence from a roundtable on AI and the agency model has stayed with me since. Someone said that AI is compressing content production from weeks to hours, and that the bottleneck has moved. It is no longer making content. It is making it matter.
I take a lot of notes at these events, pages of them, some of which I never reopen. This sentence is the one I kept turning over on the flight home, and I think it is because it reframes almost everything else I heard that week, including a thread I have been pulling on for a while now about what AI does to how we reach doctors.
Cheap content is not the win we think it is
For years the limit in pharma marketing was production. A brief took weeks, an agency was the throughput, and volume cost money. That limit is mostly gone now. Which sounds like good news, and for a day or two in the conference hallway it felt like good news. But if everyone can produce more, faster, then producing more, faster, stops being an advantage. It becomes the floor everyone stands on. The thing that used to separate a good commercial team from an average one was getting content out the door. Now the thing that separates them is knowing which content is worth making at all, and whether it actually lands once it is out there. That is a harder skill, and I am not sure pharma teams have spent the last years building it.
AI is accelerating mediocrity
There was a warning in the room that I keep coming back to, maybe because it was the most uncomfortable one. Agencies are using AI to answer briefs faster, not better. Speed without a rise in quality just produces more average work, sooner. I think this is the quiet failure of the current wave of AI adoption, inside and outside pharma. It flatters activity. A dashboard full of content produced this month looks like progress. It hides the absence of a real point of view, because nobody is asking whether the point of view was there to begin with.
The judgement problem follows AI into engagement
This is where my own thinking about AI and engagement connects, even though the conference treated them as separate topics. I have written before about the shift from omnichannel to optichannel, the move from offering a healthcare professional every channel and letting them choose, to using a model to pick the one channel and moment that fits them best. What I did not say clearly enough in that earlier piece is what this Barcelona sentence makes obvious. Optichannel only works if the judgement behind it is sound. AI can pick the channel in a fraction of a second. It cannot tell you whether the message being sent through that channel was worth sending. Point a sharper targeting engine at a confused message and you get a confused message delivered with great precision. That is not progress, it is just a more efficient way of being wrong.
The review gate is the real constraint
In pharma there is a structural reason this matters more than it would elsewhere. Every external asset has to pass medical, legal and regulatory review before anyone outside the company sees it. AI does not touch that gate. It just sends more material toward it. So the constraint everyone thought AI was removing has, in practice, moved downstream and gotten heavier. Almost nobody at the conference was talking about the operating model that can absorb a surge of content without breaking that review process. I think that is because it is an unglamorous thing to build, compared to the AI itself.
Marketers move upstream or become prompt managers
The same question of where judgement sits came up again when people talked about the marketer's own role. The old job was writing briefs, running projects, reviewing what the agency sent back. AI is absorbing a good share of that already. What is left, the part AI cannot do, is upstream: understanding the customer, shaping what the story actually is, deciding what is worth saying before anyone writes a word of it. I can see how a marketer ends up the opposite of that without noticing. Spend the day prompting a tool, hand the output to an agency, call that strategy. It looks like the upstream work. It is closer to the old job, just with a different interface.
The customer is changing at the same time
And underneath all of this the doctor on the other end is also changing, on roughly the same timeline. Within five or ten years most physicians will be people who grew up online, who would rather watch a forty second video than sit through a forty minute slide deck. I keep thinking about what that does to the field representative's job. It used to reward the person who explained things best. I wonder if it now rewards the person who listens best, which is a much harder thing to train for and a much harder thing for any AI model to fake.
So I am left with the sentence I started with, except now it feels less like a clever line from a roundtable and more like a genuine question about my own work. If the tool that produces content is no longer the constraint, and the tool that targets content is no longer the constraint, what exactly was I spending most of my time managing before this year, and was it ever the part that mattered.