Labviva, a procurement platform for life sciences R&D, has launched The Agentic Crew, an AI architecture combining specialised Small Language Models with a large language model orchestrator to automate procurement processes in regulated industries. The system features six specialist agents handling purchasing, sourcing, compliance, domain expertise, analytics and inventory management. The Boston-based company addresses limitations of general-purpose AI models, which co-founder Nicholas Rioux says are costly and prone to hallucinations in specialised contexts. Labviva's approach uses custom-trained SLMs for specific tasks, coordinated by an LLM orchestrator that integrates with existing enterprise systems. The architecture emphasises compliance through logged actions, decision trails and individual agent validation. The platform integrates with major procurement systems including Oro and SAP Ariba.

Labviva launches "The Agentic Crew": A purpose-built AI architecture to transform procurement in Regulated Industries. Specialized Small Language Models + LLM Orchestration deliver compliant, defensible, end-to-end source-to-pay automation for life sciences. June 24, 2026 10:03 ET | Source: Labviva BOSTON, June 24, 2026 (GLOBE NEWSWIRE) - Labviva, the intelligent procurement platform purpose-built for life sciences R&D, today announced the debut of The Agentic Crew, a first-of-its-kind agentic AI architecture that pairs specialized Small Language Models (SLMs) with a large language model (LLM) orchestrator to automate the full source-to-pay lifecycle for research and development organizations. The announcement, first previewed at the R&D Procurement & Sourcing in Pharma Summit 2026 on June 4, marks a decisive shift in how life sciences enterprises can deploy AI in regulated procurement environments, with compliance, auditability, and scientific accuracy built in by design. The Problem with General-Purpose AI in Pharma Procurement For years, the promise of AI in procurement has been undercut by a fundamental mismatch: general-purpose large language models were never built for the precision demands of scientific sourcing. Labviva's CTO and Co-Founder, Nicholas Rioux, has been candid about the limitations. | "When you actually run a question, if you were to distill the economics down to price per question, they're extremely high in their cost, and they have a maximum amount of hallucinations when it gets to specialty information. They're not as good at understanding complex domains that you need to be able to solve." - Nicholas Rioux, CTO & Co-Founder, Labviva | | / | Rioux's answer is not to abandon AI but to rethink its architecture entirely. Rather than relying on foundation models, Labviva builds its own small language models. Beyond the issues of cost and hallucination, Rioux points to a deeper problem: general-purpose models aren't trained on the context that scientific procurement requires. | "Another part of the problem is that the models aren't trained on the right context. Part of what we're doing is training our SLMs on the specific context relevant to our customers, which improves accuracy and lowers the cost of using agents." - Nicholas Rioux | | / | Introducing The Agentic Crew: Specialists + Orchestrator The Labviva Agentic Crew is built on a simple but powerful equation: | Specialists + Orchestrator = Compliant, Defensible Scale | | / | On one side, a suite of six purpose-built SLM specialist agents - each narrow in domain, deep in expertise, and predictable in behavior. On the other, a broad-reasoning LLM orchestrator that runs the room, coordinating across procurement tools (ERP, P2P, Orchestration), scientific tools (ELN, LIMS), and any other application via MCP. Together they form what Labviva calls the Agentic Crew: fast, efficient, on-device specialists doing the work, with an intelligent orchestrator ensuring it all coheres. THE SIX SPECIALIST AGENTS | AGENT | CAPABILITY | | Purchasing Agent | Autonomously search contracted pricing, published papers, and manufacturer specifications to find optimal buying channel based on product fit, pricing, and availability | | Sourcing Agent | Dynamically scans global marketplaces and catalogs to identify the most cost-effective, reliable suppliers across categories | | Compliance Agent | Validates every transaction against applicable Internal, Region, and EH&S Requirements before execution. | | Domain Agents | Specialty agents for chemistry, genomics, and other scientific disciplines, answering domain questions with high certainty. | | Analytics Agent[1] | Watches spend, flags trends and outliers, and answers natural-language questions with precision and provenance. | | Inventory Agent[1] | Monitors real-time stock levels, proactively triggers replenishments, and optimizes asset allocation across decentralized labs. | | / | / | [1] Not available for commercial purchase. Feature in development. "Each agent is individually validated. Together they execute compliant source-to-pay." Why the Crew Model Wins in Regulated Industries For pharmaceutical, biotech, and academic research organizations operating under strict regulations and institutional compliance frameworks, defensibility is not a feature. It is the foundation. The Labviva crew model was built with this reality in mind. Six compliance-by-design principles define the architecture: | Logged & Reviewable Every agent action is logged, traceable, and reviewable by design. | / | Decision Trail The crew model creates a decision trail at every step, not just at final output. | | Validated Specialists Specialist agents can be individually validated to internal, regional, and EHNS requirements. | / | Flagged & Re-run When one agent produces a suspect result, it is flagged and re-run without disrupting the rest of the workflow. | | Maps to Your Controls The modular architecture maps cleanly onto existing access controls, versioning, and compliance frameworks. | / | Defensible at Audit In regulated industries, defensibility is the product, not a downstream worry. | | / | / | / | A New Reality for R&D Procurement Rioux sees the broader implications as fundamental to how life sciences organizations will operate going forward. | "It's a completely new reality for the world, and that's somewhere we're looking to support. When you start to think about what makes things successful from an AI perspective, a lot of it remains the same. It's your data, it's the data quality, it's having the right access to data, it's how you do your data technical integrations." - Nicholas Rioux | | / | Labviva's agentic platform is designed to integrate natively with the orchestration layers already in place at major pharma enterprises - including platforms like Oro, SAP Ariba, and other leading P2P tools - ensuring the Agentic Crew works alongside existing investments rather than replacing them. Availability | Request a demonstration To learn more contact us through the following link labviva.com/request-a-demo | | / | ABOUT LABVIVA Labviva is the intelligent procurement platform built exclusively for life sciences R&D. Trusted by leading pharmaceutical, biotech, and academic research organizations, Labviva connects scientists with the reagents, instruments, and services they need - with the compliance, data quality, and supplier intelligence that regulated environments demand. Founded by scientists and technologists, Labviva is headquartered in Boston, MA. MEDIA CONTACT Gany Rodriguez [email protected] labviva.com