Everyone wants to use AI—and now, with the growing significance of AI agents, this is doubly true. Promising a wealth of benefits, adopting agentic AI seems like a no-brainer—but is your data management foundation prepared for its extensive demands?
Experts from Informatica joined DBTA’s webinar, Emerging Data Management Foundations for AI Agent Success, to assess the implications that adopting AI agents has on data infrastructure, offering best practices, new strategies, and essential tips. In this webinar, Josh Erhardt, director, data and AI marketing, Informatica, led the conversation with Gaurav Pathak, vice president, product management, Informatica, and Jason Du Preez, vice president, product development, Informatica.
Beginning the discussion, Erhardt pointed to the contentious statement made by Satya Nadella, CEO of Microsoft, where he argued that through AI agents, SaaS will become obsolete.
Providing some necessary context, Du Preez began by defining AI agents as “a new digital workforce—resources working to create value in the enterprise. And just like…the people alongside whom they’ll be working, agents will come in many varieties—different levels of intelligence, skill, experience, [and] they’ll come from different perspectives and even different cultures.”
And like people, these agents will need to be trained and developed to perform according to their lofty expectations. For this reason, Du Preez explained how, “I’m not sure that AI agents will replace Software-as-a-Service in the near to medium term, but they are likely to significantly impact the workforce that operates SaaS today.”
With any new technology, there is likely to be some confusion among its closely associated counterparts. AI agents, chatbots, and generative AI (GenAI) experience this phenomenon; the difference is, according to Pathak, is that AI agents are “expected to have a lot of more agency—which is defined as the capability of acting independently,” compared to GenAI and chatbots which are still backseat to its human drivers.
AI agents are the next evolution—the beginning of allowing AI more agency through innovations in reasoning and planning models, Pathak explained. The competitive advantage within this new AI agent-defined era lies within effectively utilizing proprietary data, proprietary domain knowledge, and experience, according to Du Preez.
Domain-specific data and metadata are key vectors associated with the success of AI agents, where through training, these agents will be able to utilize data to execute a variety of tasks—from coding to going after new sales prospects. “We have to train these agents for the enterprise,” emphasized Pathak. “These agents will have to be trained in some kind of simulation with human-in-the-loop annotations.”
With training in mind, high-quality datasets will be a differentiator for companies utilizing AI agents, noted Pathak. Through rigorous testing and simulations based on this data, eventually, AI agents will be “safe enough to introduce to a real environment."
Metadata, then, is exceptionally critical for agentic AI—and its importance is only growing, according to Informatica. Metadata enables:
- AI model interpretability, helping to explain how AI models arrive at their decisions and further foster trust and transparency
- Data quality and accuracy by validating and verifying data integrity—an essential component of training reliable AI models
- AI agents to personalize user experiences through analyzing patterns and preferences
Ultimately, agentic AI’s core function is powered by a flow of relevant, responsible, and robust data. Data is the gas fueling agentic AI—and data quality is of utmost importance. If data is inaccurate or incomplete, AI agents’ decisions will be flawed. This means that data governance, data quality, and data security are crucial for the successful deployment of agentic AI.
This led to the introduction of Informatica’s Intelligent Data Management Cloud (IDMC), a single platform for data and AI governance. Covering a range of functions from data cataloging to data integration and engineering, API and app integration, data quality and observability, and more, Informatica delivers a robust platform for the inventory, control, delivery, and observability of enterprise data.
This is only a snippet of the full Emerging Data Management Foundations for AI Agent Success webinar. For the full webinar, featuring more detailed examinations, a Q&A, and more, you can view an archived version of the webinar here.