Editor's Note: Agentic AI — systems that don't just answer questions but plan and execute multi-step tasks — is rapidly becoming the dominant conversation in life sciences software. As the FDA and other regulatory bodies have moved to require stronger data provenance and audit trails, the pressure on vendors to build AI that is both capable and compliant has intensified. Dotmatics' launch of Luma Agent reflects exactly that moment: a platform designed from the ground up to act on scientific data while keeping humans in control and every step traceable. For lab managers navigating an increasingly crowded AI market, the governance architecture may matter as much as the feature set.
Dotmatics has introduced Luma Agent, a new agentic AI capability embedded in its Luma Scientific Intelligence Platform. Unlike general-purpose AI tools that sit adjacent to lab workflows and generate text, Luma Agent is designed to plan and execute multi-step scientific work — analyzing data, generating reports, managing workflows, and configuring the platform itself — all triggered through plain-language instructions from a scientist or administrator.
The announcement positions Luma Agent as a direct response to a well-documented gap in the laboratory AI market: the ability to act, not just advise, while maintaining the level of traceability that regulated environments demand. That distinction matters more than it might appear on the surface.
Why agentic AI is different from AI tools you already use
Most AI tools currently deployed in life sciences labs are analytical. They answer questions about data that exists, flag anomalies, or suggest next steps. Agentic AI does something qualitatively different: it takes action. An agent can be given a goal in natural language, break that goal into a sequence of steps across multiple systems, execute those steps, and return a complete result — without a scientist manually moving data between tools or writing SQL queries.
In 2026, the focus across the industry is shifting from chatbots to agents that can execute. Life sciences companies deploying agentic AI stand to see significant improvements in clinical operations efficiency, driven by the emergence of large action models that can reason across fragmented systems like LIMS and quality management systems. The practical implication for lab managers is that the ROI calculation for AI is changing: the ceiling on productivity gains rises substantially when AI can act, not just recommend.
That shift, however, comes with a commensurate rise in risk. As organizations move toward agentic capabilities — systems that can plan, decide, and execute workflow steps within governance and compliance frameworks — stronger oversight frameworks, rigorous audit trails, and robust security controls become non-negotiable to ensure autonomous behavior remains compliant and traceable.
This is the core problem Luma Agent is engineered to address.
How Luma Agent is built — and why the data foundation matters
Dotmatics' central argument is that most AI governance failures are not model failures. They are data failures. If the underlying data is unstructured, inconsistently captured, or reconstructed after the fact, no amount of model sophistication can make the resulting AI actions verifiable.
Luma Agent is built on Luma's structured, ontology-backed scientific data — meaning data is captured and contextualized at the point of scientific work, not scraped and cleaned later. That approach is what enables Dotmatics to offer a specific governance guarantee: every step the agent takes is logged, with full tool execution traces that capture exactly what the agent did, with what inputs, and what it returned.
Lab Management Certificate
The Lab Management certificate is more than training—it’s a professional advantage.
Gain critical skills and IACET-approved CEUs that make a measurable difference.
"Scientists want more than data — they want insights, answers they can act on, trust, and trace back to the work that produced them," said Kalim Saliba, chief product officer at Dotmatics. "Because data is structured during the initial scientific work, every answer can be traced back to the queries and source data that produced it. That traceability is what gives scientists confidence in the result. We've designed Luma Agent to be able to function as a node in any AI workflow, so any external agent can call it with the full scientific context that no general-purpose tool can replicate."
The platform's connection to the Databricks infrastructure that underlies Luma also figures into the governance story. "Databricks powers Luma's ability to move from raw data to actionable intelligence at scale," said Michael Sanky, VP, Healthcare & Life Sciences GTM at Databricks. "With Luma Agent, life sciences teams can now leverage that structured foundation to not just analyze, but to act, turning insights into experiments in minutes rather than days. Luma Agent changes the equation by making complex data pipelines conversational and self-configuring."
Interested in lab tools and techniques?
Lab managers exploring the platform will want to understand the role of ontology-backed data structures in making AI outputs auditable — the underlying data model is not an implementation detail, it is the foundation that makes traceability possible at all.
What the governance gate actually looks like in 2026
Dotmatics cites a Gartner projection — drawn from Transforming R&D in Life Sciences: How AI Co-Scientists Are Accelerating Discovery (Reuben Harwood, 2026) — that 80 percent of agentic AI initiatives in healthcare and life sciences will not progress beyond initial governance checkpoints in 2026, not because the models are insufficient, but because most platforms cannot demonstrate the level of traceability and explainability that regulated environments demand.
That figure aligns with what regulators are actually doing. The FDA's deployment of agentic AI capabilities across the agency — including tools that plan, reason, and execute multi-step actions — marked a sharp departure from regulatory caution, and by mandating provenance and traceability internally, the agency has raised the bar for what compliant AI looks like in drug and device development.
The FDA expects AI systems to comply with ALCOA+ principles — a nine-attribute data integrity framework covering Attributable, Legible, Contemporaneous, Original, Accurate, Complete, Consistent, Enduring, and Available — and to ensure traceability, versioning, and immutable audit trails, with data lineage mapped from raw input to model output. Luma Agent's architecture, with its full tool execution traces and human-approval requirement before any data changes are committed, is designed with those standards in mind.
Key capabilities at a glance
Luma Agent is positioned to serve both bench scientists and informatics administrators. The capabilities span two distinct use modes:
For scientists:
- Query data in plain language without writing SQL or switching tools
- Analyze experimental results and generate reports end-to-end
- Execute calculations and hand off work to the wet lab
- Connect external language models (Claude, ChatGPT, proprietary systems) via Model Context Protocol (MCP) to query Luma's experimental context
For digital lab and informatics teams:
- Configure data schemas and workflow task types through conversation
- Set up metadata models without a specialist services engagement
- Build and iterate on the platform without opening a support ticket
- Allow external AI tools to read from and write to Luma — not just query it
| Capability | General-purpose AI tools | Luma Agent |
|---|---|---|
| Natural language queries | ✓ | ✓ |
| Multi-step task execution | Partial | ✓ |
| Audit trail / tool execution trace | Rarely | ✓ (full) |
| Human approval before data writes | Varies | Required |
| Platform self-configuration | ✗ | ✓ |
| External agent integration (MCP) | ✗ | ✓ (bidirectional) |
| Ontology-backed structured data | ✗ | ✓ |
The bidirectional MCP integration is worth noting. Luma is not just an endpoint for external agents to read from — external AI tools can also configure it, meaning a scientist using a connected assistant like Claude or a proprietary enterprise model can build schemas and set up data flows directly in Luma through that assistant, without switching context.
The Siemens context: why this announcement is bigger than one product
Dotmatics does not exist in isolation. Siemens completed its acquisition of Dotmatics — a provider of Life Sciences R&D software — for an enterprise value of $5.1 billion. With the transaction complete, Dotmatics forms part of Siemens' Digital Industries Software business, marking a significant expansion of Siemens' Product Lifecycle Management portfolio into the life sciences market.
That ownership context changes the calculus for procurement teams. Dotmatics' Luma platform now sits within a company investing heavily in digital twins and industrial AI, with stated ambitions to build an end-to-end digital thread from research through to manufacturing. Luma Agent is not a standalone product bet — it is part of a platform strategy being backed by one of the largest industrial software companies in the world, with projected medium-term revenue synergies of around $100 million annually, expected to scale to more than $500 million in the longer term.
For lab managers, that means Dotmatics is not going anywhere, and the Luma platform is likely to deepen in capability rather than stagnate. It also means integration questions — particularly around how Luma connects to manufacturing execution and quality systems — are worth asking in any vendor evaluation.
What this means for your lab informatics strategy
The broader challenge for laboratory teams right now is preparing data for the AI era — not just deploying AI on whatever data exists. As this publication has covered, structured, well-governed lab data is a prerequisite for meaningful AI outcomes, and many organizations are still operating with data environments that would make Luma Agent's governance guarantees difficult to replicate. If your lab is considering agentic AI, the first questions to ask are not about model capability — they are about whether your data infrastructure can support the traceability requirements that regulated use will demand.
Luma Agent also illustrates a broader shift that lab automation and AI strategies will need to accommodate: the line between "informatics tool" and "autonomous lab collaborator" is moving. When an AI can configure the platform it runs on, the governance and change-control frameworks that labs apply to software need to extend to the AI itself.
For informatics teams, the platform self-configuration capability is worth a close look. Administrators setting up data models and workflows through conversation — without a services engagement — could meaningfully compress implementation timelines, and it's a good candidate question to explore in any demo or proof-of-concept.
Luma Agent is currently available within the Luma Scientific Intelligence Platform. More information is available at dotmatics.com/luma/artificial-intelligence.
References
- Harwood, Reuben. Transforming R&D in Life Sciences: How AI Co-Scientists Are Accelerating Discovery. Gartner, 2026.
- USDM Life Sciences. "FDA AI Guidance 2025: What Life Sciences Must Do Now." USDM, December 18, 2025. https://usdm.com/resources/blogs/fda-ai-guidance-2025-life-sciences-compliance
- Siemens AG. "Siemens Completes Acquisition of Dotmatics." Siemens Press Release, July 1, 2025. https://news.siemens.com/en-us/dotmatics-closing/
