LAS VEGAS -- ServiceNow has agreed to acquire a data governance company that it expects to deliver data for AI agents with stronger governance and richer context.
The company, Data.world, was founded in 2015 and specializes in data catalogs and governance of data for AI. Data catalogs present a unified view of data assets in an enterprise, including where they are stored, when data was created and accessed, and how data has flowed between systems and been transformed along the way. Data.world's knowledge graph and metadata collectors will become part of ServiceNow's broader platform of data for AI, which also includes its RaptorDB and Workflow Data Fabric.
Gaurav Rewari, senior vice president and general manager of data and analytics products at ServiceNow, said in an interview during this week's Knowledge 2025 conference that all these tools represent the lifeblood of ServiceNow's AI agent platform.
"The road to agentic AI heaven, unfortunately, for a lot of enterprises, goes through data hell," Rewari said. "According to a Gartner focus group, 4% of technology leaders believe that their data is AI-ready -- that's pretty sobering. Gartner goes on to say in a separate report that by 2026, 60% of AI projects will fail because the data is not AI-ready."
The road to agentic AI heaven, unfortunately, for a lot of enterprises, goes through data hell.
Gaurav RewariSenior vice president and general manager of data and analytics products, ServiceNow
ServiceNow also already has knowledge graph support, but Rewari said Data.world brings metadata collectors and knowledge graph expertise that can enrich ServiceNow's graph further.
"This acquisition will allow us to enrich and give our AI agents more context, so that, for example, if an AI agent that's doing an HR task knows that someone is an employee who's been with a company for a certain number of years, has done a certain number of roles before, lives in this particular city, etc., then whatever recommendations it makes are going to be that much more relevant and accurate," he said.
Data.world's IP will boost users' confidence in answers they get from AI agents, according to Rewari.
"Through a combination of the data catalog and data governance, we can put rules in place and enforce them so that the quality of the data within a company is enhanced," he said. "A lot of this is about trust in data."
There are other data catalog specialists, but Rewari said Data.world's knowledge graph implementation made it a good fit for ServiceNow. In pre-acquisition research, Data.world had the highest number of users per deployment, in the high hundreds. Rewari did not name other companies considered during this research.
"It's not fun to get people to use governance tools, but Data.world made it very business user-friendly, and that attracted us," he said.
In the long term, if customers have recurring needs in the new metadata layer built with Data.world IP, ServiceNow will look to package them into insights-to-action apps prebuilt so that customers in specific industries and functional areas can deploy them out of the box, Rewari said.
AI data quality competition heats up
Configuration management databases (CMDBs) have historically been intertwined with metadata repositories, said Charles Betz, an analyst at Forrester Research. This combination and the data management concept of ontologies are getting another moment in the sun as AI agents infiltrate enterprise IT vendor platforms, according to Betz.
"Both metadata repositories and CMDBs are providing and documenting information about important IT assets, and they both come together in the enterprise architecture space," Betz said. "ServiceNow has serious credibility there and is really a serious entrant into this space, in my opinion. Other master data management and metadata management companies are going to be concerned about ServiceNow's entry here. It has every potential to be a formidable competitor."
Knowledge graphs are an increasingly hot commodity among agentic AI vendors such as Atlassian, which has fueled its Rovo Agents and AI enterprise search with its Teamwork Graph.
"ServiceNow has built from the bottom up with its base of asset and configuration items, while Atlassian is building it more middle-out based on its system of work, which is what the humans are doing with epics, stories and tickets," Betz said. "ServiceNow also has tickets but is weaker on things like user stories, although it's gradually starting to fill that gap. Atlassian is also filling the gap at the bottom end of its stack with its partnership with Lansweeper. Both realize that they need to have a complete data management portfolio."
ServiceNow's AI platform dominated discussions at Knowledge 2025 this week. Quality data is the fuel for that platform, officials said.
Preparing data for AI a work in progress
ServiceNow rolled out multiple forms of AI data quality assistance for customers over the last year. RaptorDB, introduced as a successor to MariaDB as the basis for ServiceNow's CMDB in September, is optimized for high scale and performance to support data analytics, Rewari said. A premium version of RaptorDB and an automated migration workflow can transform customers' existing databases into a more AI-friendly columnar store, and AI agents integrated with ServiceNow's Now Assist chatbot can also guide users in improving the quality of their data for AI. Third-party data management vendors such as Snowflake, Databricks and AWS have pledged certified integrations with the ServiceNow Workflow Data Fabric this week under a new partner program ServiceNow calls the Workflow Data Network.
However, ServiceNow customers that are working toward adopting its AI agents, such as medical liability insurance company ProAssurance, must still perform some human-driven internal cleanup on legacy data stores to get started.
"We put a big effort into going back through old, dirty knowledge bases and fully curating, culling, cleaning them up, structuring them in a way that then AI could go back and use them," said Max Malloy, vice president of digital experience and strategy at ProAssurance.
"[ServiceNow data quality tools] aren't going to know what is still factual, what is still accurate for us, process-wise, so that part is somewhat manual," Malloy said. "Some of it, we're choosing to just wipe out and start over, because I'd rather restructure an article so that I know it can get processed through the agentic engine more easily."
ProAssurance is training ServiceNow AI agents to work with its customer portal, based on ServiceNow Customer Central, to more quickly resolve user requests. The company plans to begin implementing AI agents in a test environment in about a month, followed by another three or four weeks of testing before moving to production.
"Oftentimes requests will be in the form of threaded emails that the insurance customer wrote their insurance agency, then that person will write to our underwriter, and our underwriter will write to our portal team," Malloy said. "We're training the AI agents to just read through those, cut through it and tell us what the person needs."
So far, email summarization in workshops with ServiceNow hasn't been perfectly accurate, but it has improved by adjusting prompts and continually training AI agents, he said. ServiceNow will likely be the company's agentic AI hub in the future, since ProAssurance has already standardized IT service management, customer service management and core business processes such as claims processing on ServiceNow tools.
"You can't have four different AI hubs, and the positioning of ServiceNow as the place where you would want that AI agent fabric to live makes sense," he said. "That's where the data is. That's where all of our processes are."
Beth Pariseau, a senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism covering DevOps. Have a tip? Email her or reach out @PariseauTT.
