From data silos to AI agents: Financial data evolution

Financial data has undergone three major phases.

I speak a bit about this on my latest podcast here.

Phase 1: Vendor-controlled silos (1990s-2010s)

In the past, getting basic earnings data meant memorizing arcane Bloomberg commands. One wrong character and you're out of luck.

This was the fortress era.

Data vendors owned the castle, controlled the moat, and if you wanted to play, you learned their commands and that was that.

Bloomberg perfected this.

The deal: We'll give you comprehensive financial data, but you adapt to our interface, learn our commands, and pay our prices.

It worked because data was genuinely scarce and expensive. Real-time feeds required satellites, massive infrastructure, and armies of people cleaning messy exchange data. Want global coverage? Pick a terminal and surrender.

Phase 2: Client-controlled marketplaces (2010s-2020s)

Then cloud platforms flipped the power dynamic overnight.

Suddenly, Snowflake, AWS, and Azure weren't just storage - they were data liberation platforms.

Instead of analysts adapting to vendor interfaces, they could demand: "Put your data where I can actually use it, or I'll find someone who will."

This wasn't just convenience. Teams could now:

• Mix datasets like Lego blocks instead of being locked into one vendor's ecosystem

• Scale analysis without mysterious terminal limits

• Integrate everything into actual workflows instead of copy-pasting between systems

• Pay for what they used instead of flat-rate terminal fees

Data vendors scrambled to adapt.

Snowflake became the poster child - suddenly smaller firms could access the same alternative datasets as large financial firms, just by spinning up a cloud instance.

But this revolution only solved half the puzzle: getting data where you want it.

You still had to do fix for the intelligence part.

Phase 3: AI-native workflows and intelligence (2020s-Present)

Now we're talking about data that actively helps you make decisions.

Instead of manually checking market data every morning, AI agents monitor unusual activity 24/7. When they spot something interesting - unusual trading volume, analyst downgrades, or earnings surprises - they don't just alert you. They research the company fundamentals, check recent news, analyze comparable situations, and deliver a complete hypothesis.

This is happening right now through technologies like Model Context Protocol (MCP).

We're moving from "data as a library" to "data as an analyst".

The difference?

In Phase 2, you pulled market data, analyzed it yourself, researched context separately, then made decisions. In Phase 3, data actively participates in generating insights.

Why this matters: The speed of insight

Alpha is getting harder to find. Markets are more efficient, the low-hanging fruit is picked clean, and traditional advantages, better data, faster computers, are commoditized.

What's left? Speed of insight.

Firms that compress research cycles from weeks to hours, or hours to minutes, win. It's not about having more data - it's about connecting dots faster and acting on those connections.

When everyone has similar information, competitive advantage goes to whoever can turn it into actionable intelligence first.

OpenBB: Built for the AI era

While traditional vendors bolt AI features onto 30-year-old terminal architectures, OpenBB started with a clean slate:

"What if we built a financial data platform specifically for the AI era?"

The result:

• No vendor lock-in: Connect to dozens of data vendors instead of being trapped in one ecosystem

• AI-first architecture: Designed for intelligent agents, not just human eyeballs

• Open ecosystem: Best innovations happen when smart people build on each other's work

OpenBB isn't replacing human analysts - it's amplifying human intelligence.

The terminal era trained us to adapt to machines. The intelligence era promises machines that adapt to us.

Don’t believe me?

Check what users are building on OpenBB in our new solutions page.