Powering LLM observability for enterprise AI with IBM watsonx

We are excited to announce four new integrations within watsonx․ai, helping organizations power LLM observability for enterprise AI. Check them out below ⤵

Fairly AI 

FAIRLY AI is proud to announce our integration with IBM watsonx.ai, combining IBM watsonx.ai's enterprise-grade generative AI capabilities with Fairly AI’s policy-aware, security-first oversight. Together, we deliver a solution built for real-world, regulated, high-risk AI deployments—bringing unmatched control, transparency, and trust to every AI system. Here's the key value proposition delivered by the integration:

• Beyond Detection: Go from "what’s wrong" to "how to fix it" and "why it matters" in your governance ecosystem
• First-Class GRC Alignment: Link risks to ISO, NIST, and OWASP controls, accelerating compliance efforts
• Automated Policy Enforcement: Codify AI policies into runtime risk management—beyond PDF-based governance
• Real-Time DevSecOps for AI: Empower product and security teams to build safer AI systems from day one

To learn more, visit → https://www.fairly.ai/integration/watsonx.

Liminal

Liminal is the most secure and flexible way organizations deploy generative AI. With Liminal, regulated enterprises are able to experience the productivity benefits of AI while enjoying world-class data protection, governance, and observability capabilities - all delivered with unparalleled cost efficiency. Through our platform ⤵

• Enterprises get unlimited secure access to all the latest and greatest models, including IBM’s watsonx foundation series
• Users can quickly and easily access their preferred models anywhere work gets done - across any site, any application, any platform
• Customizable, secure model-agnostic assistants and a ubiquitous UI mean organizations can avoid vendor lock in and are future-proofed, regardless of how LLMs and agent-based AI evolve
• Security teams enjoy the highest levels of data protection, simple yet granular administration, governance, & role-based access controls, and unprecedented observability
• Organizations are experiencing the MOST cost-effective way to securely adopt generative AI

RagMetrics 

RagMetrics provides the industry’s most trusted evaluation layer for LLM‑powered applications, seamlessly integrating with IBM watsonx.ai to help teams build, test, and monitor retrieval‑augmented systems at scale. Our platform boasts:

• Best‑in‑class LLM judgment: Our benchmarked evaluation engine matches human raters with 95% agreement—ensuring your models are scored with enterprise��grade accuracy.
• Automated hallucination detection: Instantly scan massive documents for unsupported assertions, with every hallucination flagged and the exact source quote highlighted for transparent auditing.
• Built‑in A/B testing: Compare model variants side‑by‑side through automated experiments, so you can confidently choose the highest‑performing setup.
• End‑to‑end development & monitoring: From initial prompt & retrieval tuning through continuous performance tracking, RagMetrics is your single pane of glass for managing RAG‑based workflows.
• Retrieval evaluations, not just generation: Measure the relevance and coverage of your vector stores and search pipelines—because even the smartest LLM needs the right context.

Together with IBM watsonx.ai, RagMetrics equips data science and engineering teams to accelerate time‑to‑market, mitigate risk, and drive measurable ROI from their knowledge‑driven AI initiatives. Learn more → https://ragmetrics.ai/docs.

Vellum 

Vellum.ai is an AI development platform + SDK that provides the highest reliability standards for developing, evaluating and monitoring AI-powered products in production. Designed for product and engineering teams, Vellum enables collaboration on workflow logic through both the Workflows SDK and Visual Editor.

Enterprises use Vellum because they get the core tools they need to confidently integrate AI in their products, while enabling high velocity:

• Orchestration: experiment with, debug and develop agentic workflows using the SDK or Visual Builder. Sync between the two, enabling true collaboration between your product and engineering teams.
• Evaluation & Security: In-line, Online and Offline evaluations for testing your AI workflows end-to-end.
• Release management: Decoupled AI deployments, and easy regression testing between staging and production
• Observability: Feedback loops, fully integrated with the rest of the system and metric logs to understand how your AI works in production

Check out our full list of watsonx partners here → https://www.ibm.com/watsonx/partners