What Founders and Data Leaders Should Learn From It?
Every company reaches the moment when a metric changes and no one can explain why. The data exists, the dashboards work, yet the meaning behind the number is unclear. As AI adoption accelerates, this gap becomes harder to ignore.
In a recent Grove Ventures webinar hosted by our principal Or Git, Atlan CEO Prukalpa Sankar shared what she sees inside large organizations today: intelligence is improving rapidly, but clarity still depends on context.
Prukalpa spoke about lessons from building Atlan and working closely with large enterprises. The conversation focused on what actually breaks inside organizations, what works in practice, and what both data leaders and startup founders should focus on now.
Watch the webinar here.
What Are The Key Takeaways for Data Leaders?
Data often begins with clear meaning. A Salesforce field makes sense to the sales team; a finance metric makes sense to finance.
But as data moves through pipelines, warehouses, and dashboards, the business context that explains what the number represents often disappears. What remains is technically correct but operationally confusing. This is when stakeholders begin saying they do not trust the data, not because it is wrong, but because its meaning is unclear. Restoring trust requires preserving context, not just improving accuracy.
Prukalpa shared a simple example: asking an AI analyst to identify the “top 10 customers.” Even if an organization has a single definition of customer, “top” may mean revenue to sales and product adoption to customer success. Without context, the answer may be technically correct but operationally useless.
Why does context become critical in the AI era?
Prukalpa described a shift in how organizations should think about context.
Before AI, context was often a “painkiller”, useful but not always essential. In the AI era, it becomes closer to a “pacemaker,” because for AI, context is everything.
As models improve and intelligence becomes commoditized, the key challenge becomes feeding the right context into the model’s context window. Atlan’s vision is to help organizations build a shared “world model” of their data, a structured layer of meaning that AI systems can use to produce accurate answers.
What are the four layers of context?
Importantly, organizations cannot pre-build context for every possible use case. Instead, Sankar advocates starting from high-value business workflows and engineering context around them, gradually building a reusable repository grounded in real decisions.
AI can process raw data well, but only if it understands what that data represents.
Many teams attempt to model everything before delivering value. In practice, this slows progress.
What works better is starting with high-impact workflows and engineering the context those workflows require. Over time, this creates a scalable context layer grounded in real usage.
Relevance matters more than completeness.
Sankar suggests starting with the biggest business initiatives you want to drive and working backward: identify the context required for those decisions, then build systems that capture and scale it.
Historically, an analyst’s primary value often came from technical skills like writing SQL.
That advantage is shrinking quickly.
Organizations are increasingly seeing value from analysts who deeply understand the business, anticipate questions, and bring recommendations rather than simply providing answers.
Prukalpa described the term “forward-deployed analysts” – analysts embedded close to the business. In this model, the role shifts from answering questions to anticipating them, and from delivering reports to delivering recommendations.
For years, capabilities like metadata management and lineage were often treated as hygiene work that could be postponed. In an AI environment, they become core infrastructure. AI can process raw data, but it cannot understand what that data represents unless definitions, lineage, and quality signals exist. Data loses meaning along the way in the organizationץ Organizations do not need perfect governance – but they do need enough structure for AI systems to interpret the data correctly; AI readiness is not about perfect data, but about purposeful context.
A practical observation from the webinar is that change often accelerates when something breaks. When systems fail, teams are overwhelmed, and business stakeholders stop trusting metrics, urgency appears. That is often the moment organizations become willing to invest in foundational work they previously postponed.
For data leaders, recognizing this pattern helps explain why structural improvements often happen only after trust erodes.
Watch the full webinar here.
What Are the Key Takeaways for Startup Founders?
Prukalpa sees a founder’s superpower in being able to build and understand every function in the company. By going deep into the details, not just hiring and delegating, founders gain the insight needed to hire well, spot problems early, and build stronger teams.
Models keep improving and building software keeps getting easier.
The harder challenge is enabling AI to understand how a business actually operates – through integrations, definitions, organizational knowledge, and structured context across systems.
For founders building in data and AI, the real friction often lives not in the model layer, but in the “what does this mean in this company?” layer.
Dashboards may no longer be the primary interface for insight.
Natural language access and conversational analytics are beginning to reshape how users interact with data. This shift is not only a UI change – it affects how products are designed and how insights are verified and acted upon.
Much of what matters inside organizations still lives in unstructured formats: documents, messages, text, and operational artifacts. Turning this information into usable organizational context represents a major opportunity.
Prukalpa also pointed to new possibilities across the data lifecycle, including data engineering agents that automate parts of data infrastructure.
Building software is easier than ever. But differentiation disappears quickly because capabilities are replicated fast. This shifts the founder advantage toward insight, speed, and execution. The product still matters, but the pace at which you learn and adapt becomes a core competitive advantage.
Rather than relying on durable moats, Sankar suggested thinking in terms of head starts. Distribution, deep integrations, connectors, and speed can all create meaningful advantages. These advantages compound over time even if competitors eventually catch up. In fast-moving markets, the ability to move quickly and operate in ambiguity becomes a key capability.
The future will not be purely build or purely buy.
Enterprises will build internal agents, purchase vertical solutions, and use general-purpose tools simultaneously. Founders need to understand where their product fits within this mixed ecosystem.
In some categories, companies are shifting toward outcome-based models where the value proposition is the business result rather than the tool itself.
Technology becomes an abstraction, and the company sells accountability for a specific outcome. This approach is not universal, but it is emerging in areas where implementation failure is costly.
Two Additional Operating Lessons
A central point from the conversation was that scaling requires repeatable processes, not founder-level intuition in every deal. If closing customers requires exceptional judgment at every step, the model cannot scale. The goal is to build a process that a capable operator can follow consistently.
Atlan faced this challenge early on as a horizontal platform. Many customers agreed the problem was real, but the harder question was identifying who cared enough and where the pain was intense enough to drive urgency and repeatability.
The webinar also highlighted moments where staying focused required declining opportunities or short-term revenue. Time is often the scarcest resource for a company. Opportunities that pull a team away from its core direction can cost more than they bring. Discipline around focus becomes a strategic advantage.
Don’t miss all the insights here in the webinar.
