The $35 billion question I kept getting wrong

For a long time I thought pharma's biggest commercial problem was reach. More channels, more impressions, more HCPs covered. If we could just get the message in front of enough people, the numbers would move, and every conversation I sat in about growth seemed to reinforce that framing.

I was wrong about which end of the funnel the problem actually lived at, and it took me longer than it should have to see it.

The number that sits with me now is this one. Pharma is estimated to lose around $35 billion a year in revenue because promotional content is not tailored to where a patient is in their treatment journey, or to what a specific prescriber actually needs in the moment they encounter it. That is not a reach problem, it is a relevance problem, and relevance is a much harder thing to fix than reach because it cannot be solved with a bigger budget or a louder channel mix.

I want to work through why this happens, because I think the diagnosis matters more than the treatment.

What is a pharma commercial team actually trying to do, stripped back to the essentials? Move a patient from awareness to adherence, and move an HCP from curious to prescribing to confident. Every piece of content in the stack exists to nudge one of those transitions, whether it is a brochure, a rep detail aid, a patient portal, an unbranded disease awareness site, or an HCP email campaign. Each of those pieces is a bet that a generic message will land on a non-generic human at roughly the right moment, and the bet usually loses, not because the content is bad, but because the content cannot adapt once it is published.

Static content is frozen at the moment it was approved. It does not know that the patient reading it was diagnosed three weeks ago versus three years ago, it does not know that the HCP opening the email has already prescribed in this class and is looking for comparative efficacy data rather than a mechanism of action refresher, and it cannot ask a clarifying question or remember what the person asked last time. That is not a failure of creativity, it is a limit of format.

This is where I think the industry's early swing at AI missed, and I have thought about this a lot because we swung at it too. Most pharma AI projects so far have been analytics dashboards, and while dashboards are genuinely useful, they describe the problem rather than do anything about it. A dashboard telling you that your NBA completion rate is low does not lift your NBA completion rate, it just lets you watch the number not move with more precision.

What actually changes the number is an agent that can have the conversation itself, inside the lines the team has already agreed to.

That is what we have been building at RoseRx. The agents are trained on a brand's approved content, so they stay inside the medical, legal, and regulatory lines the team already spent months getting right, but within those lines they can hold a real conversation with a patient or HCP. They adapt to what the person actually types, they remember context across sessions, they escalate when they should, and they route to the next best action that is specific to this particular person rather than to a segment of a million.

I am still careful about how much I claim. We have two top-ten global pharma companies using this in dermatology, cardiology, and urology, and the number I keep coming back to is a 310 percent improvement in completion of recommended next best actions compared to static content. That is a large number, and it is also, honestly, the thing I was most skeptical of when we first saw it in one deployment. We have now seen it repeat in others, which is the moment I started taking it seriously.

If I am honest about what this means, it means the $35 billion is not really a mystery. It is the cost of content that cannot meet a human where they are, accumulated across every interaction that would have moved someone forward and instead did not. The technology to close that gap is finally good enough to actually use in production, and the harder question is whether commercial teams are willing to rebuild around it, or whether they will bolt it on as another channel and wonder why nothing moved.

I do not have a clean answer to that yet, honestly. I think the teams that treat agentic AI as infrastructure rather than as a campaign tactic will take a disproportionate share of the $35 billion back, and the ones that treat it as a chatbot feature probably will not. I would rather be in the first group and find out I was half right, than be in the second group and find out I was late.