Reimagining the Enterprise Data Stack in the Age of Intelligent Agents (Part 1)

The future of the Enterprise data stack is unfolding before our eyes today.

This isn’t theoretical. It’s happening in real-time — inside enterprises that are beginning to challenge how decisions get made, how systems reason, and how intelligence flows through the organization.

Over the last decade and a half, I’ve had a front-row seat to how data is collected, processed, and consumed by large enterprises. And the truth is, the traditional stack — built for human control and intervention — is no longer equipped to meet the speed, complexity, or demands of modern decision-making.

What we need now is not just better infrastructure. We need intelligent architecture. In this two-part series, I want to lay out how I believe the stack is evolving — not incrementally, but fundamentally.

Part 1 focuses on the move from today’s brittle, manual data systems to an emerging intermediate model — one where intelligent agents and real-world signals begin to change how organizations see, reason, and act.

The Stack We’re Leaving Behind

Most enterprise data stacks today are made up of four brittle, disconnected layers:

1. Data (1P): Data is primarily drawn from internal systems — transaction logs, CRM activity, sensors. This tells us what’s happening inside the system, but leaves out everything that matters outside it: Where are customers going? What are competitors doing? What environmental shifts are influencing demand? These real-world signals are still mostly absent.

2. Data Features & Aggregations: Teams spend months engineering features to feed models. These features are often custom-built for narrow use cases, fragile, and difficult to scale. The process is labor-intensive and doesn’t generalize.

3. Analytics & Machine Learning: Despite all the investment in tooling, most analytics remain backward-looking. Models are static. Pipelines require constant maintenance. Dashboards tell us what happened, not what’s likely to happen next — or what to do about it. What we get out of the data is really limited by human comprehension.

4. BI & Visualization: The final interface is often a dashboard — charts that require interpretation and action. Insights and decisions become the responsibility of the individual user, not the system.

The Bridge We’re Building: The Intermediate Stack

The future of the data stack is unfolding right before our eyes. 

The stack is moving from passive infrastructure to active reasoning systems — and a new architecture is emerging. This is what I call the intermediate state. And disruption is imminent at every layer of the stack. 

1. Data

We believe first-party data will remain foundational — but on its own, it’s no longer enough. As enterprises, we increasingly need to understand what our customers do beyond our own environments: Where do they go? What captures their attention? How do macro shifts ripple into micro behaviors?

Historically, getting access to such real-world signals — consumer sentiment, people movement, local events, competitive pricing — has been fragmented, messy, and slow. But with intelligent agents, we’re changing that. We can now collect, validate, and contextualize this data continuously and at scale — not statically, but dynamically, in the moments that matter.

This marks a shift from building data lakes to building systems of data on call — where we know what to fetch, when, and why.

2. Knowledge Representations

We still work with structured and unstructured data — but the way it’s represented is fundamentally different.

Instead of engineering static features, we now create generalized knowledge representations: embeddings, vector stores, knowledge graphs. These formats are fluid, reusable, and agent-readable.

At Evam Labs , we’ve built Propheus around this idea. Our Digital Atlas captures real-world signals — mobility, demographics, sentiment, weather, pricing — and transforms them into contextual knowledge. This is no longer raw data. It’s meaning — structured for reasoning.

3. System of Agents

Once that context exists, a system of agents can sit on top of it. These agents don’t wait to be asked. They reason, simulate, infer. They look for patterns, flag anomalies, and offer decisions.

Instead of pushing data to humans and hoping for action, we give agents the ability to reason like analysts — but at scale and in real time.

This changes how data teams work too. The role of the analyst is shifting — from building pipelines to curating goals and supervising reasoning systems. The system becomes adaptive. The analyst becomes the conductor.

4. Software Interfaces

In the intermediate state, agents output through software. Human users still initiate action — but now, they’re reacting to recommendations, not just data.

The software becomes a conduit — helping humans act faster, with more precision, and greater understanding. But it’s still a transitional layer. And in many ways, it’s the last remnant of the old stack.

Why This Intermediate Layer Matters

This middle stage is the bridge between legacy and future — and it’s already being adopted by forward-thinking teams.

It’s a stack where:

• Data is not stored, but summoned

• Context is encoded into knowledge, not just stored in rows

• Agents reason, instead of humans analyzing static charts

• Software drives outcomes, not just observations.

At Evam Labs, this is how we build:

• Propheus which is building the most comprehensive knowledge representation of every place on Earth

• Poiro which is building AI Agents to bring brands closer to their consumers 

And our intelligent agents continuously learn through feedback, reinforcement, and domain-specific fine-tuning.

But this isn’t the end.

Because the future state goes even further.

It’s a world where the software layer disappears — and decisions aren’t made by humans in dashboards, but by agentic systems acting continuously, cooperatively, and autonomously.

In Part 2, I’ll dive into what I believe that future looks like — and how we’re preparing for a world where the data stack isn’t just intelligent… it’s alive.