Teva Pharmaceuticals evaluated approximately 300 technologies in one year while building an open innovation platform across its 30-factory network. Operations account for around 70% of Teva’s cost structure, which means efficiency gains in how those factories run have material impact on the company’s strategic transformation. At the Industrial AI Summit 2026, Maya Schushan-Orgad of Teva Pharmaceuticals explained how the company’s Teva Rise program selects, tests, and scales AI across a regulated pharmaceutical manufacturing network.
Why Did Teva Build an Open Innovation Platform for Manufacturing AI?
AI technology is expanding faster than any single company can build internally, and Teva is in the middle of a strategic shift called Pivot to Growth, moving from a generics-focused company to a biopharmaceutical company. That strategy has delivered 13 consecutive quarters of results. The open innovation mandate is to bring technology from outside into Teva’s operations, where 30 factories and a 70% cost share make efficiency gains material to the transformation. The platform launched challenges focused on AI in manufacturing quality and supply chain, asking operational experts to define their biggest pain points first: how to predict product quality risk earlier, how to improve asset reliability, and how to make more informed supply chain decisions.
What Is a Proof of Value and How Does It Differ from a Proof of Concept?
A proof of concept asks whether the technology works. Teva’s proof of value asks five additional questions: does it work in our environment, can it integrate and scale across a heterogeneous factory network, does it answer our challenge with measurable value, can it realistically operate in a highly regulated environment, and can users trust and act on the output? The POV process tests whether data exists at a given site, whether factory workers can incorporate recommendations into their existing workflows without excessive change management, and whether performance metrics translate into business value. Teva designs each POV in collaboration with the vendor, building decision flow mapping and traceability requirements into the system from the beginning rather than adding them after deployment. If a site lacks the data maturity or system readiness to support the technology, even the strongest vendor gets deferred until the internal environment catches up.
| Proof of Concept | Proof of Value |
| Does the technology work? | Does it work in our environment? |
| Demo performance | Integration and scale across network |
| Technical validation | Measurable business value |
| Standalone test | Workflow fit with factory users |
| Technology readiness | Regulatory and trust readiness |
Why Do Non-Pharma AI Vendors Fail Pharmaceutical Evaluations?
Across the approximately 300 technologies Teva evaluated, a clear gap emerged between vendors with pharmaceutical experience and those without. Pharma-focused vendors embed regulatory requirements, explainability, and traceability into their products from the start. Non-pharma vendors often build strong technology that lacks the compliance architecture regulated environments require. In some cases, Teva identified superior technology from non-pharma players but could not adopt it because the regulatory and trust requirements were missing. Teva extended traceability beyond compliance obligations. The data trails that satisfy regulators also feed predictive maintenance models, quality management systems, and operational improvements that take manufacturing to the next level. Non-pharma AI vendors looking to enter pharma or any highly regulated manufacturing environment need to make these characteristics core to the product, not an afterthought.
This article is based on a panel discussion at the Industrial AI Summit 2026 featuring Adam Procopio of Merck, Kristen Sauter of Adlib Software, and Maya Schushan-Orgad of Teva Pharmaceuticals, moderated by Rick Franzosa of Tech-Clarity. AI tools were used to help summarize and organize the content. Reviewed and edited by the IIoT World editorial team. The session was sponsored by Adlib Software.
