Alation launches OS for building, governing AI agents
While not the first AI governance suite, the vendor's Intelligence Operating System is somewhat unique by unifying data, context and agents in one environment.
Alation is taking on AI governance to aid enterprises attempting to build agents that can be trusted in production.
Many organizations, despite investing heavily in agentic AI development, still struggle to move AI experiments beyond the pilot stage. While myriad reasons contribute to failed experiments, including outdated technology infrastructures and lack of buy-in from executives, data-related problems are prominent among them.
In particular, the inability to connect agents with the appropriate data and business logic that gives them the context to deliver accurate outputs has hampered agentic AI development initiatives.
Alation, which already provides data catalog capabilities that enable organizations to govern data, on Tuesday launched the Alation Intelligence Operating System (AIOS).
The OS, which works in conjunction with existing Alation governance capabilities, is aimed at better enabling customers to build AI tools that can be trusted in production by connecting data, context and agents in a unified environment. Features include Agent Studio to develop agents and agentic governance capabilities, such as lineage and access controls, that automate manual work.
"AIOS is a meaningful expansion for Alation customers because it delivers needed AI governance atop the existing foundation rather than asking enterprises to start another isolated AI-governance program to manage agent data intelligence," Michael Ni, an analyst at Constellation Research, told TechTarget.
Alation's introduction of AIOS follows its unveiling of AI Governance to help organizations remain regulatory compliant as they deploy agents, outcome-based governance to govern data using agents, and CDE Manager to better oversee data related to critical operations such as reporting and risk management.
For AIOS to deliver on its promise, it needs to easily integrate with Alation's other agentic AI governance capabilities, according to Ni.
"Alation has historic strength in data discovery and governance but now has to show that its new solution areas of Agent Studio, AI governance, compliance and feedback deliver one integrated system at enterprise scale," he said.
Building for success
With the failure rate of AI projects still higher than the success rate, connecting agents with context has emerged as the dominant trend in data management and AI development. In June alone, AWS, Databricks, Microsoft and Snowflake unveiled new capabilities to connect agents with context.
AIOS is a meaningful expansion for Alation customers because it delivers needed AI governance atop the existing foundation rather than asking enterprises to start another isolated AI-governance program to manage agent data intelligence.
Michael NiAnalyst, Constellation Research
Vendors have taken different approaches, such as improving the data retrieval process, upgrading vector search and adding semantic layers. Alation, through agent-powered governance, is similarly trying to help customers build agents they can put into production.
"We developed AIOS from a mix of customer demand and a fundamental shift we observed in how AI is redefining data management," Satyen Sangani, Alation's co-founder and CEO, told TechTarget. "AI provides so much power to do the work that historically has been side-of-desk, things like stewardship, quality, lineage. [Additionally], the result of this work is now so much more critical and impactful."
In particular, the work is needed to catch incorrect outputs that agents are confident are correct, he continued.
"When software breaks, it throws an error, but when an AI agent breaks, it produces a confident, incorrect answer, and most organizations have no system of record to catch it," Sangani said. "AIOS closes that gap … so enterprises can trust the AI they're already running instead of finding out after the fact."
Data-related problems that frequently stall agentic AI projects include bad data causing an agent to act on improper context, agents misreading context that isn't clear, and agent drift with instructions, tools and training degrading over time and causing the agent to diverge from its intended goal.
Alation's new operating system is designed to address each problem by providing the following capabilities:
Agent Studio, an environment aimed at enabling users to develop trustworthy agents that call on governed proprietary data and business logic for context.
Governed data and AI workflows to ensure that agents remain in compliance with changing regulations.
Control over context, access and lineage so that each agent-powered decision can be traced and certified.
Conversational analytics so users can query and analyze governed data.
"The AIOS represents a notable shift toward automation by using agents to proactively curate metadata, map data lineage and scale enterprise governance," William McKnight, president of McKnight Consulting, told TechTarget. "It's a clever way to prevent confident fails [by agents, but] introduces some vendor lock-in for the sake of faster AI deployment."
By pre-certifying metadata and enforcing governance policies, AIOS appears appropriately constructed to reduce agent hallucinations, he added.
Meanwhile, AIOS is in step with current AI governance trends rather than pacing the market, according to McKnight.
"The entire enterprise data governance industry has simultaneously pivoted to launch specialized suites for governing autonomous AI agents," he said. "They are all racing to solve the exact same enterprise problem -- ensuring that autonomous AI does not act on bad data or violate compliance boundaries."
Ni likewise noted that by connecting AIOS to Alation's catalog and other governance capabilities, the new operating system seems logically put together to aid customers as they build and manage agents. However, for it to truly operate as intended, AIOS needs to not only help govern agents under development but also once they're deployed, he continued.
"AIOS has the foundational components," Ni said. "The open question is how far governance extends into production execution. A complete offering should continuously monitor agent behavior, propagate the correct user identity and entitlements, enforce policies at the point of action, and provide a runtime audit trail."
From a competitive perspective, Ni noted that Actian, Atlan, Credo AI, IBM, Microsoft and ServiceNow are all addressing agent governance. However, Alation is among the first to unify capabilities such as context, lineage and conversational analysis in an open architecture.
"The real test will be if context advantage can demonstrably translate into stronger runtime decisions and fewer operational failures," Ni said.
Plotting the future
With AIOS now generally available, Alation is preparing to release new ontology capabilities – governed definitions of data and relationships similar to semantic models – and improving its existing lineage, data quality, data observability and AI governance tools, according to Sangani.
Following the introduction of AIOS, Ni noted that there's an opportunity for Alation to turn the new OS into a context layer for agents working across environments such as databases, multiple cloud data platforms and an array of SaaS applications.
"Alation's opportunity is to connect what the enterprise knows with what its agents are doing right now, while opening a much larger market among AI platform and operations teams," he said.
McKnight, meanwhile, suggested that Alation could better serve existing customers and perhaps attract new ones by integrating AIOS with data observability capabilities and advancing its agentic AI tools to include more autonomous insight generation and automation.
"Furthermore, expanding the platform to support unstructured and multimodal data is considered crucial for building a comprehensive AI ecosystem," he said.
Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management.
