Chalice AI and PurpleLab today set out a partnership that puts longitudinal medical and pharmacy claims records behind custom predictive models for healthcare advertisers, with activation running across connected television, social platforms and the open internet. The two companies published the agreement through Business Wire at 9:00 AM Eastern Daylight Time on September 14, 2026, from New York. No client names, pilot results, pricing or availability date accompanied it.

In Short

Two companies have agreed to plug a large pool of anonymised medical and pharmacy billing records into software that builds custom prediction models for advertisers. Drug brands and healthcare marketers would use those models to guess which people and which doctors are worth reaching, instead of buying ready-made audience lists someone else defined. Nothing about how well it works has been published yet, so the announcement describes a capability rather than a result.

What the two companies are combining

PurpleLab describes itself as a healthcare analytics company powering real-world data covering more than 330 million United States patient lives and upwards of 3 million healthcare professionals, according to the announcement. Chalice AI describes itself as a platform-independent AI media decisioning company whose custom models can be deployed in all major demand-side and supply-side platforms, as well as Meta and YouTube.

The stated mechanism is a division of labour rather than a merged product. PurpleLab supplies longitudinal medical and pharmacy claims records. Chalice supplies the modelling and optimisation layer that converts those records into predictions, and the deployment path that carries the predictions into buying environments. Together, according to the companies, they will build bespoke, brand-specific predictive models designed to identify new audience members likely to be interested in specific drugs or therapeutic areas, uncover prescriber opportunities, predict adoption and attrition, and drive incremental growth.

Four distinct use cases sit inside that sentence, and they are not equivalent in difficulty. Predicting which consumers might be receptive to a therapeutic category is a patient-side modelling problem. Uncovering prescriber opportunity is a provider-side problem, resting on records describing clinicians in their professional capacity. Adoption and attrition modelling requires longitudinal continuity in the claims file, not just breadth. Growth that is genuinely incremental requires a counterfactual, which no audience model supplies on its own.

PurpleLab's scale figures are not new to the record. The Video Advertising Bureau added the company to its Advanced Measurement Solutions Directory on July 21, 2026, where its HealthNexus platform was described as combining more than 14 billion annual medical claims, more than 3 million healthcare providers and more than 330 million United States patient lives. That directory entry also recorded a social media measurement product delivered through data clean room infrastructure with Decentriq, and an eligibility alerts feature offering four to seven days of pre-visit intent signal. Those figures were vendor-supplied rather than independently audited, a caveat the directory itself carried.

Outcome-predicted audiences against static segments

The argument Chalice puts forward is directed at the audience segment as a unit of purchase. Campaigns often begin with a predefined audience or a broad proxy for intent, according to the companies, leaving marketers optimising against assumptions rather than the outcome they want to drive.

Rather than relying solely on static audience segments, Chalice creates audiences based on their predicted likelihood of an advertiser-defined outcome. PurpleLab's healthcare intelligence provides the underlying signals needed to understand patient trends and healthcare professional behaviours, while Chalice translates those signals into predictive insights that can inform media decisioning and activation, according to the announcement.

Model refresh is the operational claim attached to this. Models can be refreshed as new healthcare signals and campaign performance data become available, the companies state, which in principle closes a loop that segment-based buying leaves open. Claims data, however, arrives on a lag. Medical and pharmacy claims move through adjudication before they settle into an analytics file, and the interval between a clinical event and its appearance in a longitudinal dataset is measured in weeks rather than hours. Neither company published a refresh cadence, a claims latency figure, or a description of how campaign performance data is joined back to the healthcare signal.

Adam Heimlich, CEO of Chalice AI, framed the pitch around dissatisfaction with packaged audiences. "Healthcare marketers don't need another off-the-shelf audience. They need the ability to start with the business or patient outcome they're trying to achieve and build an AI strategy specifically around that objective," he said in the announcement. "The opportunity is to turn trusted healthcare intelligence into a predictive advantage unique to each brand, then put that intelligence to work wherever media is bought. PurpleLab gives us the foundation to do exactly that."

The compliance perimeter around healthcare signals

Healthcare marketers have access to increasingly sophisticated clinical and claims intelligence, but translating those signals into scalable media decisions remains difficult, particularly in a highly regulated category where privacy, security and compliance considerations shape how data can be used, according to the companies.

Transparency, security and data provenance are described as central to the partnership. Models are developed in privacy-first, secure environments, with marketers retaining visibility into the data categories and provenance informing their models and the methodology used to translate those signals into media decisions, the announcement states.

Kyle Armstrong, VP, Agency and Platform Sales at PurpleLab, tied the value of the modelling to the defensibility of the inputs. "AI is only as valuable as the data powering it and the confidence marketers have in how that data is being used," he said. "By bringing PurpleLab's real-world healthcare intelligence together with Chalice's custom predictive decisioning, we're giving marketers a more transparent and actionable way to connect healthcare signals to the outcomes they are ultimately trying to drive."

What the announcement does not do is name the regulatory framework it is working within. HIPAA is not mentioned, nor is de-identification methodology, nor whether models are trained inside a clean room, nor whether a business associate agreement structure applies. The distinction between patient-side and provider-side targeting carries real legal weight in this category: National Provider Identifier records describe clinicians professionally rather than patients clinically, which is why provider targeting has expanded faster than patient targeting across healthcare ad tech. StackAdapt brought self-serve NPI targeting to pharmaceutical advertisers on January 6, 2026 with impression-level reporting it described as compliant. DeepIntent opened its Helix infrastructure to third parties on March 19, 2026, covering what it said were more than 3.7 million providers and over 240 million patient lives.

The Chalice and PurpleLab announcement describes both patient and healthcare professional use cases without separating the compliance treatment of each. Marketers evaluating the offer would have to establish that boundary through diligence rather than from the published material.

Chalice's route into the supply side

The company arriving at healthcare data is not a healthcare specialist. Chalice has spent three years building a position in containerised decisioning, the practice of running a buyer's model as code inside an exchange rather than inside a demand-side platform.

Zillow became the first brand advertiser to pilot containerised real-time bidding in August 2025, running Chalice models inside Index Exchange infrastructure and surfacing page-level signals such as ad ratio per page and refresh rates that ordinary bid requests do not carry. IAB Tech Lab released the Agentic Real-Time Framework for public comment in November 2025, standardising how containerised agents participate in auction infrastructure, with Chalice among the named contributors to ARTF. Bedrock Platform ran the first full containerised DSP bidder inside Index Cloud on April 21, 2026, with Chalice listed among partners building on the same platform.

The pattern accelerated through mid-2026. PubMatic opened its auction to partner models through Decision Fabric on June 1, 2026, with Chalice in onboarding. Chalice models began running inside OpenX infrastructure for Hyundai, reported on June 16, 2026, an arrangement that gave the advertiser visibility into a wider pool of supply than a demand-side platform is typically rationed. Magnite named Chalice among four partners in its containerisation position paper on August 6, 2026, without attaching a latency ceiling or a case study to the claim.

According to Chalice, the company pioneered the first containerised bidding agent in an SSP in 2024, enabling advertisers to optimise supply around their goals. That history matters for reading the PurpleLab agreement. The distribution layer already exists across multiple exchanges and two walled gardens; what the partnership supplies is a category-specific signal to feed it.

A category absorbing infrastructure at speed

Healthcare advertising technology has taken on a dense build-out over the past twenty months, and most of it has run through a small number of specialist platforms rather than general-purpose ones.

epocrates opened clinician workflow inventory to programmatic buying through DeepIntent on May 26, 2026, the first integration of in-workflow clinical reference inventory into a demand-side platform. DeepIntent added Vistar Media out-of-home and point-of-care inventory to its DSP in August 2026. CheckedUp won Media Rating Council accreditation for waiting room television delivery in June 2026, bringing independent audit into a channel that had operated without one. DeepIntent put a 50% planning time reduction figure on its Cora AI product on September 11, 2026.

What separates the Chalice arrangement from those is the absence of a dedicated healthcare demand-side platform in the middle. DeepIntent, StackAdapt and Doceree sell healthcare targeting inside their own buying environments. Chalice sells a model that runs inside whichever environment the advertiser already uses, including exchanges where the buying decision happens on the sell side. For a pharmaceutical brand with existing seats across several platforms, that is a different procurement question than adding another specialist DSP.

Measurement is where the difference gets tested. Script lift, the difference in prescription volume between an exposed group and a matched control, remains the currency pharmaceutical media is graded in, and neither company said how models built on PurpleLab signal would be evaluated against it. PurpleLab supplies data into script lift reporting elsewhere in the market, which makes the omission notable rather than fatal.

Where the money has been moving

The commercial backdrop is a category whose budgets have been shifting on regulatory pressure rather than media innovation. Prescription drug brands raised multiscreen television spending 53% to $4.95 billion between October 2025 and March 2026, against $3.23 billion a year earlier, according to Video Advertising Bureau analysis of Nielsen Ad Intel data. The Food and Drug Administration had sent enforcement letters over deceptive drug advertising on September 9, 2025, concentrated on digital and social placements. Brands responded by consolidating into television, a channel with decades of established precedent.

Policy moved in the other direction on the programmatic side. Google dropped certification requirements for Authorized Buyers promoting prescription drugs from January 2026, having introduced restricted drug term certification for personalised targeting in July 2025. The result is a market where the same advertiser faces different gates depending on the buying route.

Against that, the activation surface Chalice names is telling. Connected television, social platforms and the open internet are the three places pharmaceutical budgets have been concentrating, and the first two are where identity resolution is hardest to audit from outside.

What the announcement leaves unstated

Several gaps sit in the published material, and they are the kind that determine whether the capability converts into spend.

No launch or availability date is given. No pilot advertiser is named. No match rate is disclosed between PurpleLab's patient and provider records and the identity infrastructure used to activate across connected television, social platforms and the open internet. No performance figure accompanies the claims, which stands in contrast to how Chalice positioned earlier partnerships, where a named advertiser and a cost reduction figure carried the release.

The privacy language is descriptive rather than technical. Privacy-first, secure environments is a category description, not an architecture. Marketers retaining visibility into data categories and provenance is a governance promise whose implementation is unspecified. Neither company stated whether model training occurs inside a clean room, under what aggregation thresholds outputs leave that environment, or how patient-side and provider-side signals are separated at the modelling stage.

Designed to support a growing range of patient and healthcare professional use cases, the collaboration can help marketers identify likely new patients and prescribers, model adoption and attrition, and connect predictive insights to media and creative strategy, according to the companies. Models can be activated through Chalice across connected television, social platforms and the open internet, allowing healthcare marketers to apply the same custom intelligence across channels.

That is the full extent of the operational description. Everything measurable about it remains outside the published record.

Timeline

Summary

Who: Chalice AI, a platform-independent AI media decisioning company led by chief executive Adam Heimlich, and PurpleLab, a healthcare analytics company represented in the announcement by Kyle Armstrong, VP, Agency and Platform Sales. The intended buyers are pharmaceutical, life sciences and healthcare marketers and their agencies.

What: A partnership combining PurpleLab's longitudinal medical and pharmacy claims data with Chalice's custom AI modelling, producing brand-specific predictive models that identify potential patients and prescribers, forecast adoption and attrition, and inform media and creative decisions. Activation runs through Chalice across connected television, social platforms and the open internet. No client, pricing, availability date or performance figure was published.

When: Published through Business Wire at 9:00 AM Eastern Daylight Time on September 14, 2026. It follows PurpleLab's entry into the Video Advertising Bureau measurement directory on July 21, 2026 and a sequence of Chalice containerisation deployments running from August 2025 through August 2026.

Where: Issued from New York. The underlying data covers United States patient lives and United States healthcare professionals, and the activation channels named are connected television, social platforms and the open internet.

Why: Healthcare marketers hold increasingly detailed clinical and claims intelligence but face difficulty converting it into media decisions inside a category constrained by privacy, security and compliance requirements, according to the companies. Campaigns built on predefined audiences optimise against assumptions rather than outcomes, and the partnership positions outcome-predicted audiences and refreshable models as the alternative.