Data Products, AI Agents, and the Future of Work: Insights from Sanjeev Mohan

In a fast-evolving digital landscape, data is no longer just a byproduct of business operations—it’s becoming the engine of innovation. 

In a recent episode of Data Radicals, Sanjeev Mohan, former Gartner analyst and now principal at SanjMo, joined host Satyen Sangani to unpack the buzz around data products, AI agents, and what lies ahead for enterprise data management. The conversation illuminated the growing importance of treating data not as infrastructure, but as productized, trustworthy, and intelligent assets.

Here are the top takeaways from this rich and forward-looking discussion.

What is a data product, really?

While the term “data product” has gained traction in recent years, its definition remains elusive. Mohan explains that confusion arises because people interpret it through the lens of their own roles—business leaders, engineers, and data scientists all see it differently.

At its core, a data product is “a reusable and standardized data asset that delivers some measurable value,” says Mohan. But beyond the asset itself, what truly distinguishes a data product is its ownership, usability, and trustworthiness.

A data product builds trust in data, Mohan notes. It’s a shift from ad hoc querying to a product management mindset—with APIs, versioning, discoverability, and designated owners.

Whereas traditional data pipelines were often treated as one-off projects, data products introduce a lifecycle mentality. They evolve, improve, and—crucially—get retired when they no longer deliver value. Think of them like software applications, versioned and maintained over time.

Why data products matter

So why go through the effort of turning data into products? The answer lies in scalability and accountability.

Modern enterprises are overwhelmed by the demand for trusted data. Centralized data teams simply can’t keep up. Mohan references the philosophy of data mesh, originally championed by Zhamak Dehghani, which decentralizes data ownership to domain teams. This shift empowers those who understand the data best—typically the business units—to produce and maintain their own data products.

Mohan turns to the world of healthcare to demonstrate the value of granting data management responsibilities to those who know the data best.

“Let's say I'm a hotshot data engineer and somebody tells me, here's your clinical data, and I want you to take this clinical data and derive these reports for me,” he says “I don't know what those clinical codes mean at all! I'm not a trained clinician. 

But if you give that responsibility to the business team and tell them you are responsible, it's your data, you curate it, because you understand the meaning and produce a well-defined, reusable asset, then that is a problem we are trying to solve in a data product.”

Ultimately, data products aim to eliminate ambiguity and inefficiency. When something breaks—like a dashboard showing incorrect figures—users should know exactly who owns that data and how to get support. That’s the promise of the data-product model: clear ownership, discoverability, reusability, and embedded trust.

Data products and AI agents: A perfect match

One of the episode’s most powerful insights is how data products can serve as the foundational layer for AI initiatives, particularly AI agents.

AI agents, according to Mohan, are the next evolution of AI automation. Built on top of large language models (LLMs), these agents can sense, reason, plan, and act. Think of them as hyper-intelligent digital coworkers who can summarize emails, monitor competitor activity, or manage operational workflows.

Yet, if AI agents are going to reason by leveraging internal data, they need context, trust, and semantics – all of which data products provide. 

Mohan elaborates, about a year and a half ago, “a lot of models were hallucinating like crazy.” He believes that data products can deliver that trust layer that AI so desperately requires. “It's curated, metadata is published, APIs are published, there's a data contract. What if I put my agents, my assistant, my chatbots on top of the data product? Then I can improve reliability and trust in my outcome,” he shares.

Sangani adds, “It also is the case that the data product can then travel with its own semantics and descriptions that would give the LLM confidence and context through which to talk to it.”

In other words, running AI on uncurated, raw data is risky. But if your AI sits on top of a curated, well-documented data product, it’s safer and far more reliable. In this model, data products act as trust layers for AI, complete with metadata, contracts, and usage semantics. This setup empowers AI agents to return accurate, contextualized answers—whether they’re interacting with dashboards, APIs, or natural language prompts.

How AI agents are transforming jobs

Mohan argues that early iterations of AI agents will be role-based digital assistants that automate repetitive cognitive tasks. Examples range from reading and prioritizing emails to conducting competitive research.

These agents are not replacing jobs, Mohan argued, but transforming them. “Every job will change, but new jobs will be created,” he argues, going on to point out that having an AI agent is like having 10 subject matter experts doing the research for you.

From a strategic standpoint, organizations that fail to embrace AI agents may find themselves left behind. “Some organizations may not even have a choice. Because if your competition is doing it, then why would you wait? You'd be forced into this space, in my opinion.”

Predictions for the future of data and AI

Looking ahead, Mohan sees the coming year as one of rapid transformation. He envisions the emergence of an entirely new category of software: the Agent Management System. By 2026, he predicts it will be so prevalent that Gartner will include it in a Magic Quadrant.

Mohan also sees the unification of structured and unstructured data as a major trend. Technologies like Apache Iceberg and embedding models are making it easier to work across formats. Semantic layers and metadata management will be key to unlocking this potential. 

Consolidation is likely on the horizon. Mohan argues that leaders don’t want different stacks for AI and analytics – it’s all just data – and the AI agent be the compute engine that sits on top of unified, governed data. 

As for governance, Mohan sees a twofold opportunity: infusing AI into traditional governance tools, and governing AI itself. The former will enhance discoverability and data quality. The latter will help enterprises monitor AI models for cost, performance, and hallucination risk.

Final thoughts: Experiment boldly, govern wisely

If there’s one theme that runs through the entire episode, it’s the power of experimentation. Mohan encourages organizations not to wait for perfect definitions or industry standards. Instead, they should test, learn, and iterate—whether with data products, AI agents, or new architectural models.

“Unless you experiment, how do you know?” Mohan says. 

In an era where AI and data are reshaping the future of work, this advice couldn’t be more relevant. Companies that build a strong foundation of trusted, reusable data—and harness it with intelligent agents—will not only survive, but thrive.

For more insights on how to get your data AI-ready, check out alation.com/ai-ready.