SPEAKER:
Andrea Goudie, Head of Strategy and Partnerships, Ostro
KEY TAKEAWAYS:
- 30% of patients now trust AI for health information pharma refuses to engage with conversationally
- 87% of patients lack the literacy to understand standard pharma branded content
- Production data shows 100% MLR/PRC approval rates across conservative regulatory environments
- HCP AI adoption at 66% daily use has already outpaced pharma's engagement model
- Open-text patient queries represent a compounding first-party intelligence asset pharma is forfeiting
"30% of US consumers, patients and caregivers, say they are getting their health information from AI," Andrea Goudie, Head of Strategy and Partnerships at Ostro, argued at a pharma digital marketing conference in Philadelphia, "and they are trusting this information as a source of truth more than ever." That data point would be alarming enough on its own. Paired with what follows, it becomes an indictment. "87% of those same patients still lack the reading comprehension skills to understand the average piece of pharma content on a pharma branded site." These are not two separate problems with two separate remedies. They are the same problem from opposite ends: pharma's content is inaccessible to the patients it's designed to reach, and those patients have found something more accessible. The migration is not a failure of patient education. It is a structural consequence of an engagement model that was never designed to answer questions, only to make statements. Goudie's session, delivered weeks after the announcement of Ostro's acquisition by Veeva, pressed a direct challenge: the compliance rationale pharma has used to justify one-way communication may no longer be protecting anyone.
Pharma Built a Library When the World Wanted a Search Bar
The mismatch Goudie describes is not a design flaw that better UX can correct. Every standard pharma digital channel; branded websites, email sequences, SMS programs, is architected to terminate engagement rather than sustain it. A patient arrives with a question and leaves with a PDF. An HCP receives a detail piece calibrated to a persona that doesn't match their patient population. The channel ends. What makes this structurally damaging rather than merely frustrating is the behavioral training effect: each dead end doesn't just lose an interaction, it teaches the user where not to go next time.
Goudie crystallized the architectural failure in a single observation: "We think about pharma as still building a reading world when the world is really asking questions." The content production model that pharma has optimized for decades; rigorous, reviewed, one-directional, was built on the assumption that users want to consume information. That assumption has been invalidated by user behavior, and the gap between what pharma produces and what patients and HCPs actually do is now measurable.
On the HCP side, the adoption velocity data makes the urgency concrete. With "an 80% increase year over year in HCPs who are using AI every single day in their practices" and "66% of them today using it every day," this is not an emerging behavior to monitor, it is current-state reality that pharma's engagement model is already behind. Goudie drew the EHR adoption parallel to frame the trajectory, but that comparison actually understates the case. EHR adoption was driven by institutional mandates, reimbursement incentives, and regulatory requirements. AI adoption among HCPs is voluntary and demand-driven, which means pharma cannot wait for the wave to crest. There is no policy lever that slows it.
Volume fatigue and relevance fatigue are different problems, and pharma's content strategy has consistently conflated them. "Greater than 60% of HCPs still say they are overwhelmed by pharma content," Goudie noted. "They believe that pharma sends them too much information, and it is not personalized for them." The response to HCP overwhelm has typically been editorial, cut the frequency, sharpen the targeting, improve the creative. But if the underlying channel cannot respond to what an HCP actually needs to know in a given moment, the volume problem is a symptom, not the disease.
The compounding dynamic is what makes the current moment strategically different from previous cycles of digital channel hand-wringing. Every time a patient or HCP hits a one-way wall on a pharma property and migrates to an AI tool instead, two things happen simultaneously: pharma loses a data signal about what that user actually wanted, and the AI tool gains one. The competitive position doesn't hold steady while pharma deliberates, it deteriorates with each failed interaction.
The Compliance Excuse Has an Expiration Date
"In our industry, our excuse, if you will, for why we have not adapted in many ways is almost always compliance and regulatory concerns." The word "excuse" is doing deliberate work. In an industry where compliance functions as both genuine constraint and organizational identity, recharacterizing it as an excuse is not rhetorical provocation for its own sake, it is a direct challenge to the logic that has justified decades of one-way infrastructure investment. Goudie is not dismissing regulatory risk. She is arguing that regulatory risk has been used to foreclose conversations that the evidence suggests are approvable.
The compliance objection to conversational engagement rests on three connected claims: that open-ended patient and HCP input cannot be controlled, that real-time responses cannot be reviewed and approved, and that adverse event monitoring at conversational scale is operationally untenable. The non-generative architecture Goudie describes addresses all three. Responses are drawn exclusively from pre-approved content libraries. Nothing is generated de novo. What MLR and PRC review is the library, not each individual exchange, which is structurally identical to the review process for every other content asset pharma already produces.
The proof is not theoretical. "Our solution has 100% MLR/PRC approval records at our clients, even among those who proudly are the most conservative from a regulatory and compliance perspective." A claim like this, made publicly at an industry conference, is either verifiable or it isn't. What it forecloses is the argument that compliance approval is inherently incompatible with conversational engagement. If the most conservative regulatory environments in the industry have signed off, what remains of the objection is not risk, it is inertia. Inertia is a leadership problem, not a regulatory one.
The open-text input question is where the compliance conversation typically breaks down, and where Goudie's reframe is most pointed. "Open text, while it might give some of you heartburn, the two words 'open text,' has really fundamentally changed how we are able to adapt now in the age of AI. It also generates a just unlimited, tremendous amount of user insights." The instinct to close off open-text input,to replace it with structured menus and constrained navigation, is understandable as a risk-minimization reflex. But that reflex forfeits something significant: a direct, unmediated pipeline into what patients and HCPs actually want to know.
This is where the strategic implication extends beyond what Goudie explicitly frames. As third-party cookie deprecation reshapes targeting capability across the industry and syndicated data sources face growing limitations, organizations that have been systematically collecting structured first-party queries will hold an intelligence advantage that compounds over time. Open-text inputs don't just improve individual interaction quality, they constitute a proprietary dataset of real questions, real information gaps, and real decision barriers that no purchased data panel can replicate. For organizations that have roadmapped conversational engagement as a 2025–2027 initiative on the assumption that compliance resolution requires multi-year effort, Goudie's evidence compresses that timeline to weeks. The strategic question is no longer whether this is approvable. It is what each quarter of delay is actually costing.
Discover more on this topic at Pharma Customer Engagement USA 2026 (October 27-28, Philadelphia) - where commercial, marketing, medical, data and AI pioneers converge. Explore the agenda here.
Qualified Leads Are Worth More Than Educated Ones
"For the fellow millennials in the room, we don't pick up the phone, and we're not going to call your patient assistance program." The default escalation architecture in pharma patient services; the 1-800 number, the nurse hotline, the live chat that routes to a call center, was designed for a communication preference that is generationally eroding. This is not a UX inconvenience. It is a structural failure in the escalation model itself, one that becomes more pronounced as the patient populations pharma most needs to reach skew younger and more digitally native.
The most common organizational anxiety about deploying conversational AI, that it will displace field reps and patient support staff, inverts the actual value proposition. A patient who arrives at a rep interaction already educated on mechanism of action, already past the basic access questions, already prepared to discuss their specific clinical situation, is a fundamentally different engagement than a patient starting from zero. The economics of field time change when AI handles the foundational layer. Rep capacity shifts from baseline education to conversion-ready clinical dialogue.
Goudie's deployment data indicates that clients see "at minimum about a three times faster access to resources and a lot less frustration among HCPs and patients",but the commercial significance of that figure is not primarily a user experience metric. It is a cost-per-meaningful-interaction reduction for field operations, and it compounds across a territory. The repositioning is precise: "We don't mean to say that the reps and field teams are no longer relevant or that we would replace them... Instead, let's make better use of their time and send more qualified, more prepared patients and HCPs there." The field force doesn't shrink. Its inputs improve.
The Calculus Has Inverted
For roughly a decade, the perceived risk of conversational engagement, compliance exposure, adverse event monitoring burden, brand safety at scale, exceeded the perceived risk of maintaining one-way channels. The static website, the gated PDF, the email that links to a contact form: these were not failures of imagination. They were rational responses to a risk environment where the downside of a compliance violation was visible and the downside of an inaccessible patient experience was diffuse and hard to attribute.
That calculus has inverted. Goudie's deployment data indicates that Ostro's technology is now live across more than half, and likely closer to 70%, of the top 20 pharma companies, with additional penetration across midsize firms, medical affairs functions, patient services, and ex-US markets. Early movers have already made the calculation. The window for competitive differentiation through conversational engagement is narrowing; the window for competitive disadvantage through inaction is open and widening.
What the open-text insight points toward requires a step beyond Goudie's own framing to fully articulate: a first-party data moat with compounding returns. Organizations that deploy conversational AI now don't just improve patient and HCP experience in the current quarter. They begin accumulating a proprietary dataset of real questions, real information gaps, and real decision barriers that no syndicated research, no CRM database, and no third-party data purchase can replicate. Over 12 to 18 months, that intelligence reshapes not only digital content strategy but clinical messaging architecture, medical affairs priority-setting, and field force deployment logic. The organizations that wait don't simply fall behind on channel engagement,they fall behind on understanding what their customers actually need, and that gap is harder to close than the technology gap.
The compliance question has been answered. What pharma leadership now faces is whether it can afford the intelligence deficit that widens every month the answer goes unacted on, in an environment where the AI tools filling that vacuum answer to no MLR committee, no adverse event protocol, and no brand standard whatsoever.
To get you highlights of Pharma USA 2026 faster, we are using generative AI technology to summarise the transcripts of the sessions. If you have any feedback about the summary, please contact lucy.fisher@thomsonreuters.com.
Discover more on this topic at Pharma Customer Engagement USA 2026 (October 27-28, Philadelphia) - where commercial, marketing, medical, data and AI pioneers converge. Explore the agenda here.
