Reuters Pharma USA 2026: Lessons from 70+ Sessions

This year, Reuters Pharma USA drew 1,200+ leaders across 100% of the top 50 pharma companies with teams ranging from analytics and commercial operations to market access, medical affairs, and patient services. A lot of ground was covered during two days and more than 70 sessions, workshops, and roundtables. What follows is my attempt at pulling together key themes and takeaways for data-minded pharma teams heading into the second half of 2026.

The format was four parallel tracks — Customer Engagement, Data & Infrastructure, Launch & Value, Evidence & Outcomes — plus keynotes, workshops, and roundtables. Sessions used silent disco-style headsets and AI-powered note-taking tools that provided real-time summaries. It was clear this wasn’t a conference about whether to adopt AI. It was a conference about what to do now that everyone already has.

What’s Keeping Pharma Teams Up at Night

Many of the problems people are struggling with aren’t about technology. They’re about fragmentation, trust, speed, and the distance between patients and the systems designed to serve them.

The patient access gap is worse than the numbers suggest

ZS’s opening keynote included a stat that set the tone for the conference: “44% of patients are saying, hey, this is too hard, too costly, too challenging, and it’s not worth it to me.” Nearly half of patients are opting out of the system entirely.

Lilly USA’s President added another dimension. While most chronic diseases get 90–95% coverage and reimbursement, obesity medications reach only about half of commercial lives. That gap isn’t just a payer issue — it’s a structural failure in how the system treats certain diseases.

And the friction is absurdly granular. ZS shared that 30% of specialty prescriptions were being canceled because pharmacies called patients from unknown numbers. Patients didn’t pick up. Prescription canceled. That’s not a data problem. That’s a caller ID problem.

Analytics teams are buried in the wrong work

Regeneron’s Executive Director of Commercial Strategy described the reality bluntly: “It sometimes takes multiple hours and days before you can simply answer questions. And sometimes when you answer the question, it is already too late. People have moved on and they’ve already made a decision.”

Analytics teams want to be strategic. Instead, they’re fielding ad hoc requests all day. The volume of incoming questions crowds out the time needed to find the answers that actually matter.

Meanwhile, data recency emerged as a recurring gap across the Data & Infrastructure track. Panelists from Novartis, BMS, and others noted that their data is often weeks or months behind the decisions it’s supposed to inform. And even when data arrives on time, the assumptions behind it can be wrong. One panel pointed out that claims data reflects reimbursement, not physician adoption — a distinction that can quietly break entire analytic models.

Organizational silos are the real AI blocker

Every conference talks about silos. This one made the cost specific. Multiple sessions pointed to the same structural problem: different teams pursuing AI independently, using different vendors, building different pipelines, and setting up separate governance processes. GenerativeX estimated that large pharma companies easily have 30–50 AI pilots running simultaneously with no coordination between them.

ZS called the result for what it is: “We’ve followed the strategy of letting a thousand flowers bloom, which has resulted in a thousand graveyards of successes. No pilots fail and yet few pilots scale.”

Brand plans are part of the problem too. As one session put it: marketing, medical, and access teams come together, align on brand pillars, go back to their own drawing boards, create their plans, and never shall they meet again.

“We’ve followed the strategy of letting a thousand flowers bloom, which has resulted in a thousand graveyards of successes. No pilots fail and yet few pilots scale.”— Managing Partner, ZS

Trust — not accuracy — is what’s stalling AI

If there was a single word that defined the conference, it was trust. Not accuracy. Not performance. Trust.

Regeneron put it simply: “Trust is number one. You don’t want hallucinations coming through where you ask a different variation of the same question and it gives you a different answer.”

A panel featuring leaders from Gilead, Sanofi, and Pfizer took it further: “The world doesn’t change at the speed of technology. It changes at the speed of trust.” And then one panelist corrected the framing entirely: “I don’t think it’s correlated to trust. I think it IS trust. The best model that doesn’t get used is useless.”

What’s Actually Working

Amid the challenges, several sessions showcased ideas that have moved past the pilot stage and into real commercial impact.

Lilly Direct: bypassing the system at scale

Eli Lilly’s direct-to-patient platform now serves over 1 million people through self-pay, bypassing traditional PBM and insurance friction entirely. Combined with their Employer Connect program and the upcoming launch of Orphaglipron — Lilly is trying to prove that the access problem is solvable if you’re willing to rethink the distribution model.

Genentech’s evidence-based formulary

Genentech described a simple but radical shift of basing their internal formulary on evidence instead of rebates, with the rationale “if there’s more evidence for a medicine, our employees pay less out of pocket.” The projected savings equate to millions of dollars over several years. The point isn’t the dollar figure — it’s that an evidence-based model can replace the rebate-driven one.

Predictive field engagement at Sanofi

Sanofi shared results from deploying predictive analytics across their field reimbursement team: a double-digit increase in customer interactions plus higher-quality conversations. The shift was behavioral, not just technological. Reps stopped reacting and started going into meetings with proactive intelligence about prior authorizations and copay support access.

GSK’s AI-powered MLR

GSK brought AI into their Medical Legal Review process and reported saving over hundreds of hours of review time. The application handles claims checking, referencing, brand accuracy, and asset routing. “Technology can now reduce risk while increasing speed” — a sentence you don’t hear often about MLR!

Five Themes That Defined the Conference

1. The decision window is shrinking

I heard this from Takeda, AstraZeneca, Regeneron, Merck — different stages, same pressure. IRA has compressed commercial timelines. One panel noted that if the exclusivity period drops to nine years, companies need to hit peak performance within three. Another panelist from Inizio Ignite said it plainly: “We don’t have the one to two years of grace anymore. You’re almost in a series of sequential hyper-launches.”

The organizations pulling ahead aren’t just moving faster. They’ve made sure that when an answer surfaces, everyone trusts it enough to act on it. ZS quantified it: 50% faster decision cycles are correlated with a 10% lift in brand performance.

2. The pilot graveyard isn’t a technology problem — it’s a trust problem

This theme ran through almost every AI-related session. Leaders don’t just want AI that’s accurate. They want AI they can defend to their CFO, their compliance team, their board. And accuracy alone doesn’t get you there. The AI has to know who’s asking and what the question means in their specific commercial reality. That’s the difference between a demo and a deployment.

Genentech offered a great analogy: “Many of our AI solutions are smart, capable, and seemingly engaging, but without any memory or context. Every morning, like Lucy in 50 First Dates, we start with zero.” They traced considerable manual burden to this institutional amnesia — teams can’t find pre-approved assets, reviewers re-learn content performance every time.

“The world doesn’t change at the speed of technology. It changes at the speed of trust.”— said during panel session featuring Gilead, Sanofi, and Pfizer

3. Field teams are still walking in blind

PharmaForceIQ described the field experience as “a black box to us. They send things over the fence. We know we’re supposed to execute against this alert, but we don’t really know what it means.”

The Sanofi team captured the root cause: “Field teams don’t need more data. The problem isn’t data — there is too much noise.” Their fix was specific: instead of “recommended topics — efficacy and dosing,” reps get messages like “Dr. Smith has engaged heavily with our efficacy content in the last 7 days. Recommend focusing on dosing because 40% of HCPs who wrote their script engaged with dosing information.”

4. Pull is giving way to push

The quiet commercial shift that surfaced across multiple sessions: stop querying, start receiving. Instead of analysts pulling reports, organizations are building always-on monitors that detect changes and deliver daily digests. Less query interface, more co-pilot.

BioMarin shared their evolution of moving from quarterly segmentation cycles to daily patient leads. And Merck flagged that lower-latency data sources like hub referrals and copay redemptions can arrive daily, making real-time push models practical now, not theoretical.

5. The human-in-the-loop is being repositioned, not removed

AI handles the repeatable. Humans stay for judgment, relationships, and strategy. The organizations getting this right are redirecting teams toward work that actually requires a human. That cultural change is harder than the tech.

One panel used a historical analogy that landed: “When they first brought electricity in, it was lift and shift. They traded out steam engines for electric motors but left everything else the same. It wasn’t until 20 years later that leaders reimagined the entire system.” Most pharma organizations are still in the lift-and-shift phase with AI. The real gains come from redesigning the system around it.

What Teams Can Do Now

The conference was full of practical, implementable advice. Speakers didn’t just describe problems — they offered specific actions. Here’s a curated set organized by what you could do today, and 30/60/90 days out.

This Week

• Audit your vendor phone numbers. 30% of specialty prescriptions were canceled because pharmacies called from unknown numbers. Check yours on Monday.
• Map your AI landscape. Inventory all AI tools across your team, then compare with colleagues in other functions. You’ll find redundancy.
• Run a deprioritization exercise. Get your cross-functional team together. Identify 1–2 things to stop doing, 1–2 to double down on.
• Check your claims data assumptions. Claims data reflects reimbursement, not physician adoption. If your analytic engine treats it as adoption, flag it.

30 Days

• Add “what, why, why now” to your NBA engine. If your next-best-action system just says “recommended topics: efficacy,” reps will ignore it. Rewrite suggestions to include the specific insight, the rationale, and the urgency. This is a configuration change, not a rebuild.
• Shift KPIs from awareness to utility. Replace awareness and engagement metrics with speed to first appointment, speed to diagnosis, and time to first fill.
• Engage compliance before you have funding. Bring legal, compliance, and regulatory into AI conversations now. Existing frameworks may already cover your use case.

60 Days

• Launch an AI ambassador program. Takeda shared this during their launch case study: 5–10 field reps who are natural AI adopters teach their peers. Higher adoption than top-down mandates.
• Build or refresh a patient panel. 15 people, intentionally diverse. Run creative concepts past them before launch. UCB, AstraZeneca, and MyHealthTeam all emphasized this as critical for rare disease launches.
• Set up payer’s-own-data partnerships. Approach 1–2 payers about using their patient data for evidence generation. When you’re working with the payer’s own data and their own patient population, the evidence is uniquely defensible.

90 Days

• Build a semantic layer for conversational analytics. Regeneron stressed the importance of unified framework that connects business vocabulary to data so users can self-serve instead of filing tickets with analytics teams. Decision cycles go from days to minutes.
• Create a Market Access Analytics Center of Excellence. Crosswalk PBM portal data, formulary policy, and prescription data. Position the COE as augmenting pricing and contracting teams, not replacing them.
• Bring AI into MLR review. AI can handle claims checking, referencing, brand accuracy, and asset routing. If you’re not at least piloting this, you’re leaving time on the table.

The filter for everything above

ZS offered a question that’s worth taping to the wall:

“Is this reducing distance to the patient, or just decorating the existing system?”— Managing Partner, ZS

Apply it to every initiative. If the answer is “decorating,” deprioritize it.

If Your Analytics Team Is Buried in Ad Hoc Requests

The themes above — shrinking decision windows, trust gaps, field teams flying blind, the shift from pull to push — all trace back to one structural problem: the people who understand the data can’t get answers out fast enough, and the people who need the answers can’t self-serve.

That’s the problem Tellius solves. If you’re heading into H2 trying to figure out how to free your analytics team from the request queue and get trusted answers into the hands of the people making decisions, let’s talk →

‍

‍

‍