PUNE: As companies around the world race to adopt artificial intelligence, many believe the latest AI models will determine who succeeds. However, enterprise AI and CPQ leader Eshaan Jain believes the real advantage lies elsewhere — in the quality of the data that companies already possess.Addressing industry leaders during a virtual interaction, Jain said that the success of agentic AI systems depends more on the quality of company data than on the AI model itself.Speaking about the growing excitement around AI agents in business, Jain said many companies are focusing on the wrong thing.“Companies want to drop an agent on top of quote-to-cash (Q2C) and expect it to close deals,” Jain said. “The agent is only as good as the contract, product, and pricing data underneath it. If that data is messy, the agent guesses. A guessing agent inside a revenue workflow is a liability, not an assistant.”Jain, who leads Salesforce and Vlocity CPQ transformation for T-Mobile in Bellevue, Washington, explained that AI agents work by using information already stored inside enterprise systems. According to him, clean and structured data is what allows AI to make reliable decisions.“Agentforce can read a customer’s history, build a quote, and route an approval without a human touching it,” he said. “It does that by reading your existing CPQ rules, your product catalog, your clause library. Those are the systems people dismiss as legacy and want to rip out. Those systems are the moat.”Sharing lessons from his time at Amazon, where he worked on machine learning systems that analyzed contracts, Jain told industry leaders that preparing data was often more important than building the AI model itself.“Everyone remembers the model. The real effort went into cleaning and labeling years of contract language, so the model had something honest to learn from,” Jain said. “Skip that step and you get a confident system that is wrong at scale.”He advised future engineers to understand the importance of data quality before deploying AI systems.“The unglamorous work nobody wants to fund is the work that decides the outcome,” he said. “Clause extraction, catalog cleanup, quote-data quality. Do it first and the AI program has a chance. Skip it and you have funded a demo, not a result.”Jain also highlighted the future challenges of AI, especially the cost of running large models and their environmental impact.“Right now, teams measure whether the agent works,” he said. “Soon they will measure what each agent decision costs in dollars and in CO2. The programs that survive past 2027 are the ones that designed for that on day one.”When asked what skills engineering students should develop for careers in AI, Jain encouraged them to build knowledge across multiple domains.“The people who lead this shift can talk to a CFO about margin and to an engineer about data models in the same meeting,” Jain said. “That range is rare, and it is what the agentic enterprise actually needs.”Ending the session, Jain stressed that strong data foundations will determine which companies succeed with AI in the coming years.“Treat your CPQ and contract data as a revenue product, with real ownership and real investment,” Jain said. “Do that, and agentic AI pays off. Treat it as a configuration job, and no amount of AI will save the quarter.”
