Ataccama automates data quality with agentic AI-fueled suite

Ensuring trust

Trusted data is crucial to developing AI applications and analytics tools.

Ataccama One Agentic represents a significant upgrade over previous versions of Ataccama's data trust platform. The key advancement is the introduction of the One AI Agent, which transforms the platform from a traditional rule-based system requiring manual configuration to an autonomous system.

With generative AI (GenAI) and agents able to make workers better informed and more efficient, many enterprises have increased their investments in AI development since ChatGPT's November 2022 launch by OpenAI marked significant improvement in GenAI technology. In response, data management vendors en masse have created environments within their platforms designed to simplify using proprietary data to train AI tools so they can understand the unique characteristics of an organization.

But despite rising interest in AI development coupled with platforms built to simplify building agents and other AI tools, an overwhelming majority of AI projects -- estimated as high as 80% -- never make it into production.

Numerous factors play into the failure rate. One of the main ones is data quality, or lack thereof. Without high-quality data that leads to accurate outcomes, AI projects are doomed to fail.

Ataccama's tools are designed to help users ensure data quality. Now, the vendor is automating the process with agentic AI-powered automation capabilities aimed at making it faster and easier to build the trust in data required to develop tools that inform business decisions.

A combination of customer feedback and observing market trends provided Ataccama with the impetus for developing Ataccama One Agentic, according to Jay Limburn, the vendor's chief product officer.

"Customers were spending too much time fixing data problems manually, from writing rules and reconciling reports to debugging pipelines," he said. "At the same time, the rise of AI made it clear that enterprises need a way to trust the data driving those systems."

In addition to making trained data experts more efficient, a benefit of Ataccama One Agentic could be that it enables non-expert users to work more extensively with their organization's data, according to Timm Grosser, an analyst at BARC U.S.

By automating processes, such as defining rules, and by generating and testing complex data expressions from natural language prompts, he noted that business users will be able to perform tasks on their own that previously had to be done by data engineers and other data workers.

"These capabilities bring business users closer to data quality processes while reducing reliance on technical teams," Grosser said. "[Natural language] features … significantly improve onboarding, troubleshooting and productivity. Combined with optimizers that simplify language and automate routine tasks, it delivers a more intuitive user experience and accelerates everyday data management."

In addition to the One AI Agent, Ataccama One Agentic features:

• A Model Context Protocol (MCP) server that lets agents securely access governed data in Ataccama and external sources, such as large language models.
• A Data Trust Index to measure dataset reliability.
• Reference Data Management to keep vital information, such as customer segments and regulatory categories, consistent across pipelines.
• Continuous data observability to detect and resolve data quality issues.
• Ataccama's current data quality engine.

By combining the One AI Agent with pre-existing features, Ataccama One Agentic is reasonably put together to accomplish its intended purpose, according to Catanzano.

"It appears logically constructed to achieve its automation goals based on its multi-layered architecture," he said.

Grosser, meanwhile, noted that Ataccama is not the only vendor to provide agentic AI capabilities that address data quality. For example, Monte Carlo features agents for data observability while Alation's platform includes agents that automate data catalog capabilities, such as governance and discovery.

However, Ataccama might differentiate itself by offering Reference Data Management, according to Grosser.

"That will help organizations ground their models in consistent business terminology as they integrate AI with their proprietary operational processes," he said.

Informa TechTarget

While lack of trusted data is one reason many AI projects fail, another is fear that turning processes over to autonomous AI tools could lead to mistakes humans would not make that cause organizational harm.

Ataccama One Agentic itself is an AI-powered platform that removes humans from certain processes. However, Limburn noted that while an agent does the mechanical work of addressing data quality, people set the intent and review the outcomes before data is used to train models and inform applications.

"It's a collaboration between humans and automation, where the agent keeps data quality running in the background so teams can focus on analysis, policy and strategy rather than maintenance," he said.

Catanzano similarly pointed out that by documenting each step for review and including continuous observability to detect and resolve issues before they affect models and applications, Ataccama One Agentic's design should allay fears about turning processes over to AI.

"The platform appears to address potential enterprise concerns about automated data governance through several safeguards," he said.

Next steps

With Ataccama One Agentic now GA, the vendor's product development plans include adding more automation through the One AI Agent and expanding the new platform's capabilities to handle unstructured data, according to Limburn.

"Our focus is on making Ataccama the data trust layer for enterprise AI to ensure every model, copilot and workflow operates on data that can be explained, audited and trusted," he said.

Catanzano, meanwhile, suggested that Ataccama grow beyond data quality and governance by developing agents for other areas of data management.

"Ataccama could expand its agentic approach into predictive data management, having the AI agent anticipate data needs based on business patterns and automatically prepare datasets before they're requested," he said.

In addition, Ataccama could add industry-specific agent-based workflows and improve its MCP server's integration capabilities, Catanzano continued.

"An opportunity lies in expanding the MCP server's integration capabilities to work with a broader ecosystem of AI tools and enterprise applications, potentially becoming the de facto standard for trusted data access across multi-agent AI environments," he said.

Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than 25 years of experience. He covers analytics and data management.