A Data Product Leader's Guide to Winning in the Agentic Era

If you ask the market, the outlook for data businesses is bleak. Valuations for information services businesses are down 15–20% in the past three months. Many critics argue that AI will make it easy for a few enterprising engineers to build datasets that took incumbents years to assemble. But data companies are not software companies. The same force threatening to disrupt SaaS — the ability to generate code instantly — also fundamentally reshapes the economics of using data. An analysis that once required a PhD now only needs a few well-designed prompts and the right agentic infrastructure. That means the market for data can expand significantly.

The line between winners and losers will be drawn in the next 12–18 months. The winners will embrace AI — not only as a new workflow or destination for their data, but as a fundamental shift in how they build products. The losers will look inward to protect their existing assets and let the biggest technological wave in the past century pass them by.

“AI will not lift all boats equally. Info services winners will be those who double down on their core assets while integrating AI into their data streams and workflows.” — Nathan Graf, Evercore

“Agent-ready” is not enough

Most data companies recognize that AI will change how their products are consumed. Data product teams see how customers are using AI tools to interact with their data: whether that's building lists, analyzing results or even writing SQL. Most teams are responding by investing in ways to bring their data into these new workflows, making products “agent-ready.” That typically means one of two things: releasing a Model Context Protocol (MCP) that lets agents interact with existing APIs, or offering semantic data shares that include data and context in one package.

That's necessary, but not sufficient to win. In these cases, all of the intelligence — the very thing that makes AI powerful — is driven by another service. The key decisions about how to navigate your data products, what fields to use, how to match against other fields, and how to interpret results are all left to the user's agent.

What is an “agentic data product?”

To unlock the full power of agents, data companies need to go beyond making data available in AI workflows. They need to build AI-native capabilities that bring their deep domain expertise to bear. That means a new type of data product: one that draws on both a provider's unique knowledge of its datasets and decades of experience turning that data into value for customers.

An agentic data product is a composable set of services, wrapped around a data asset, that makes data accessible and actionable by an AI agent while maintaining the governance and control your business requires. Agentic data products don't just let external agents interact with existing capabilities like an MCP — they let teams build new agentic capabilities on top. Platforms that enable teams to build and deploy these capabilities quickly will define the next generation of data infrastructure.

What makes agentic data products unique?

The big difference between agent-ready and agentic data products is intelligence. An agent-ready data product has no intelligence layer — it provides the raw materials for an external intelligence layer (e.g., a customer's agent) to make inferences effectively. An agentic data product adds three layers on top of raw access:

• Access. A secure, governed way for a user or agent to query the “gold” or “silver” datasets that power all of your services — essential because it unlocks the agent's ability to do what agents do best: reason over data and generate answers on demand.
• Context. Semantics, tools and skills that capture everything you know about your data and how it's best used, so the agent can navigate your data efficiently and accurately.
• Intelligence. An LLM with an “agent harness” that allows your customers' agents to interact collaboratively with your data. Instead of relying only on the underlying data and context, the customer's agent can ask broader questions and collaborate with your agent to generate answers — iterating with follow-up questions until it reaches the right result.

Across these layers, the difference between a traditional data product, an agent-ready data product and an agentic data product comes down to how much of the work — access, context and intelligence — the provider has already done for the customer. A traditional product hands over raw access and expects the customer to build everything else; an agent-ready product adds machine-readable context (semantic models, MCP) but still leaves reasoning to the customer's own agent; an agentic data product adds a provider-controlled agent that can reason, collaborate and generate answers directly.

How agentic data products unlock value for data companies

Agentic data products unlock an entirely new set of use cases that allow data providers to radically accelerate their own internal operations and begin to provide more services to customers that were typically out of reach.

Internal operations:

• Automated data quality. Use agents to continuously validate, profile and flag anomalies across your datasets — replacing brittle, rule-based QA pipelines with intelligent checks that adapt as schemas evolve and catch issues static rules miss.
• Scaled sales & customer success. Equip sales and customer success teams with agents that answer complex data questions on demand — reducing time-to-resolution, improving onboarding, and supporting more customers without scaling headcount linearly.

New products:

• Conversational analytics. Deploy conversational analytics that let non-technical users explore, enrich and analyze data in natural language — generating new revenue streams by dramatically improving users' ability to generate insights.
• Agentic collaboration. Build a new set of capabilities that collaborate with customer agents to deliver complex analyses using both customer and provider data.

A new infrastructure for the agentic age

To capture this opportunity, data product teams need a new generation of capabilities: tooling that can transform the raw, general intelligence of AI models into domain experts on their sprawling — often unwieldy — data estates, and do so without going through yet another multi-year re-platforming project many are just now finishing from the cloud. That's why agentic data product platforms like Bobsled are essential investments for any data company.

Data platforms remain your factory floor: managing the core pipelines that aggregate and refine your data. Data product platforms connect seamlessly with your data platforms and are where you manage the context and intelligence required to turn “silver” and “gold” data into customer-ready agentic data products.

Data product platforms should deliver:

• New capabilities, not just new workflows. An integrated intelligence layer that delivers insights, not just data; powerful APIs and built-in MCPs to integrate anywhere; and a white-labeled conversational interface to get started fast.
• Customer-ready accuracy & reliability. Intelligent context management that learns over time, programmable skills that refine analysis, and world-class, multimodal agents designed for complex analysis.
• Faster time-to-market. Pre-built integrations to every data platform, instant deployment via integrated MCPs, APIs and apps, and agent orchestration tuned for sophisticated analytics.
• Commercial-grade governance. Granular entitlements that make it easy to deploy for each customer, real-time observability to see the questions customers ask, and a foundation pressure-tested by the world's most security-sensitive teams.

The opportunity ahead

The question this paper opened with — apocalypse or panacea? — is a false one. The real question is whether data companies remain providers of datasets, or evolve into providers of intelligence. The answer depends entirely on what data companies do next. Moats around data access are eroding, and the pace will only accelerate. But for data companies that build agentic data products — embedding their domain expertise into the context and intelligence layers agents rely on — AI doesn't just preserve the business, it transforms it.

Agentic data products have the chance to bring scale to the data business in unprecedented ways:

• Expand your addressable market. Collapse the skill barrier and get your data in the hands of every decision-maker — not just analysts and data scientists.
• Sell answers, not just access. Move up the value chain from data feeds to intelligent, on-demand analyses that command a fundamentally different price.
• Build compounding defensibility. Embed your domain expertise into the context and intelligence layers, where it compounds with every customer interaction — creating a moat that gets stronger with AI, not weaker.