Artificial intelligence could automate as much as 50% of enterprise data management work within the next 12-18 months as companies increasingly deploy AI agents to handle tasks ranging from data discovery and integration, to governance and quality management, according to a senior Salesforce executive.“We can definitely go from 20% to 50% of data management work getting automated, with humans in the loop,” Gaurav Pathak, senior vice president of product management at Salesforce, told ET on the sidelines of Informatica World, the annual conference of Salesforce-owned Informatica in Las Vegas. “We're already seeing a large number of manual processes being replaced. Pepsi talked about how it wants AI to take on more such activities while allowing employees to focus on more strategic work.”Informatica, a US-based enterprise data management software company, helps organisations integrate, govern, catalogue, and manage data across on-premise and cloud environments. The company has positioned itself as a key provider of the data infrastructure required to power AI applications.Pathak said Informatica employs about 2,400 people in India across Bengaluru, Hyderabad, and Chennai, and plans to continue expanding its research and development footprint in the country as it builds more AI-led products.In May last year, Salesforce acquired Informatica for around $8 billion in a cash deal, seeking to strengthen its AI and data capabilities amid growing demand for enterprise AI agents. Salesforce said Informatica's data management platform will help improve the quality, governance, and accessibility of data used by its AI offerings.However, with Indian enterprises remaining relatively cautious about moving AI projects from pilots to production, Pathak pointed out that the biggest hurdles include concerns around business outcomes, governance, privacy, and data quality.“If you're creating a support agent on your website, you want to see the number of cases come down and customer satisfaction go up,” he said. “Without context, without all of that data, you cannot bring a support agent into production.”He added that poor data remains a major risk for AI deployments. “We used to say ‘garbage in, garbage out’. With AI agents, it's ‘garbage in, gospel out’. They take in all the garbage, but because they are very persuasive, they don't even let the user know that they have data quality challenges.”During the Informatica World 2026 event, the firm also announced partnerships with AI cloud platforms Databricks and Snowflake to help companies operationalise agentic AI systems using governed enterprise data.Pathak said enterprises are also grappling with skills shortages around evaluating and governing AI systems. "We still need humans in the loop to provide the right guidance, the right guardrails, and the right standards to the agents," he said.(The correspondent was in Las Vegas at the invite of Informatica)