Teva and Factory AI: Choosing with Proof of Value, Not Proof of Concept

On November 4, 2026, IIoT World published a summary of a panel from the Industrial AI Summit 2026 in which Maya Schushan-Orgad of Teva Pharmaceuticals described how the company chooses artificial intelligence technologies for its manufacturing network. On stage with her were Adam Procopio (Merck) and Kristen Sauter (Adlib Software); the moderator was Rick Franzosa (Tech-Clarity). The session was sponsored by Adlib.

The context: 30 plants and about 300 technologies

Teva's open innovation platform is called Teva Rise and spans a network of 30 plants. About 300 technologies were evaluated in a single year. Operations account for roughly 70% of the cost structure, so efficiency is central to the "Pivot to Growth" strategy, the shift from generics to biopharma, which according to the article has delivered 13 consecutive quarters of results.

The starting point was not technology but problems: Teva asked operations experts what their biggest pain points were. Examples cited include predicting product quality risk earlier, improving asset reliability, and making better-informed supply chain decisions.

Proof of value: five questions beyond "does it work?"

A proof of concept asks whether the technology works. Teva's proof of value adds further questions:

The five questions, in brief:

• Does it work in our environment?
• Can it integrate and scale across a heterogeneous network of plants?
• Does it deliver measurable value against the problem?
• Can it operate in a highly regulated context?
• Can users trust the output and act on it?

The process also checks whether the data actually exists at a given site, whether operators can use the recommendations without heavy change management, and whether technical metrics translate into business value. Each proof of value is designed together with the vendor, with decision flows mapped and traceability requirements included from the start rather than added after release.

Data maturity and the vendor gap

If a site lacks the necessary data or systems maturity, even an excellent vendor is put on hold until the site is ready. A gap also emerges: vendors with pharma experience build in regulatory requirements, explainability and traceability, while those from other industries often have solid technology but lack the compliance architecture. In some cases Teva found superior technology outside pharma but could not adopt it.

Traceability trails, moreover, are not only for compliance: according to the article, they feed predictive maintenance models, quality management systems and operational improvement initiatives.

Why it matters for those leading industrial projects

The message is that whether an AI tool reaches production is often decided by data readiness at each site and built-in traceability, more than by model quality. Teva's framework is repeatable and tests integration, data availability, workflow fit and user trust.

Not just "does it work?", but "does it work here, does it scale, does it deliver value, does it withstand regulation, and can it be trusted?"

Practical takeaway and caveats

Teams building or buying industrial software can use the five questions as acceptance criteria: data availability per site, integration testing across different systems, metrics tied to value, audit requirements defined up front, and confirmation that the output is understandable to those who must act on it.

Keep in mind that this is a summary of a vendor-sponsored panel, and that the article also mentions the use of AI tools to summarize the content, with editor review. It provides no quantitative results, such as savings or the number of tools adopted, nor does it say how many of the roughly 300 technologies passed selection. It is therefore an interesting method, not proof of results.