The Evolution of Financial Intelligence
- Daryl Bowden
- Jun 9
- 5 min read
The Next Great Mutation in Financial Services I joined the Merrill Lynch dealing floor in London in 1996 by 1999 there were more than 150 traders, sales traders, and support staff on the equity floor alone—each part of a human network pushing, negotiating, interpreting, and executing trades across Europe’s capital markets.
It was loud. It was energetic. It was expensive. It was full of high iQ staff with usually immense cognitive capacity. By 2008, the game had changed forever.
A combination of direct market access (DMA), algorithmic execution, and exchange connectivity compressed execution commissions from 60 basis points down to 7. Entire careers disappeared. Headcount collapsed. What had once been considered a human art was re-engineered into machine logic—run in code, executed in microseconds, and costed at fractions of a cent.
That story didn’t just unfold at Merrill Lynch. It happened across the sell side and the buy side. From the global bulge brackets to the local brokers, the pattern was the same: low-touch replaced high-touch, and the industry reshaped itself around speed, connectivity, and automation.
Today, we stand at the edge of an almost identical transformation.
But this time, it isn’t about execution. It’s about intelligence.
The Great Compression: A 30-Year Transformation in Execution
To understand what comes next, it’s worth looking back.
For decades, trading was a human process—relationship-driven, voice-based, filled with nuance and judgment. Commission rates were sticky, trades were chunky, and access was intermediated by institutions that held the keys to liquidity and price discovery.
The first cracks appeared with electronic order books in the 1990s. Then came DMA, which allowed buy-side desks to connect directly to exchanges. And then came algorithms—volume-weighted average price (VWAP), percentage of volume (POV), arrival price strategies—designed to fragment, pace, and execute trades without intervention.
By the mid-2000s, a new reality had emerged. Trading wasn’t just digital—it was automated, optimized, and dehumanized.
And commissions followed suit. From 90bps in the 1990s, average execution costs in large cap equities fell to single digits. Sales traders became execution consultants. Human traders gave way to quants and coders. What used to take 150 people now takes 15—or fewer.
The logic of that collapse was simple:
[Connectivity + Code + Compute = Compression]
Now, that same logic is coming for financial intelligence.
Intelligence Will Be Next
Today, most people still think of artificial intelligence in financial services as a narrow tool: something to help with repetitive workflows, compliance checks, risk assessments, or summarising documents. That’s not surprising. Consultants love these kinds of pain points. They’re low risk, low cost, and politically safe.
But that is not what the next decade is about.
This is not about solving small problems.
This is about rethinking the entire premise of intermediation in financial services.
Just as electronic execution made traders redundant, AI-powered intelligence will make large swathes of the current research, advisory, and insight value-chain obsolete—not by replacing individuals one-for-one, but by building pipelines of knowledge that are:
deeper
faster
cheaper
and fundamentally more scalable than any human model ever could be.
Let me say this clearly: The next wave of transformation will not focus on a single pain point.
It will create an entirely new distribution layer for intelligence.
The Toolkit Is Now in Place
There is a reason this is happening now—not 10 years ago, not 10 years from now.
The toolkit is complete.
Cloud: Virtually infinite compute and storage. Scalable, global, cheap.
Compute: GPUs, TPUs, edge and parallel processing. Real-time analytics.
Architecture: Microservices, APIs, stream processing, and graph databases.
AI: Not just LLMs, but agentic models that reason, plan, and act.
Together, these components mean we can now build pipes of knowledge in the same way we built pipes for trading execution.
Think about it.
An agent can parse 10,000 filings, analyse 2,000 earnings transcripts, cross-reference macroeconomic indicators, and update you on material changes across 10,000 stocks before breakfast. It doesn’t sleep. It doesn’t forget. It doesn’t suffer cognitive overload. And it gets better every week.
This is not a chatbot.
This is not a dashboard.
This is an intelligence layer.
And it is coming.
Agents as the New Analysts
What happened to execution is already happening in intelligence.
Where once you had teams of researchers covering sectors and themes, you now have multi-asset, multilingual, 24/7 agents trained to interpret, assess, and surface the most relevant signals—contextually, affordably, and instantly.
This won’t just be a co-pilot model for retail. It will underpin entire firms.
Agents will know more than any individual can.
They will track thousands of signals.
They will learn across domains.
They will scale at near-zero marginal cost.
In time, these agents will become interconnected, drawing from structured data, unstructured text, graphs, audio, and even video. They will not wait for prompts. They will act based on triggers. They will alert. They will write. They will strategise.
And just like the Merrill traders of the 90s, the traditional models of research and advisory will not survive it unchanged.
A Pipe of Intelligence, Not a Patch for Workflows
Let’s be honest. Most enterprise AI today is palliative. It’s designed to ease pain, not create transformation. It focuses on small tasks because that derisks delivery. It sells well in pilot phases. It avoids political friction.
But transformation does not come from fixing spreadsheets.
It comes from challenging who has the power to know.
When DMA took hold, it shifted power from sell-side traders to buy-side desks. When agents become the norm, power will shift again—from institutions that hoard insight, to platforms that democratise it.
Financial intelligence will no longer be scarce, slow, or gated by expensive intermediaries.
It will be:
Accessible
Available
Affordable
This will not happen overnight. But it will happen.
And once it starts, it will not stop.
Why Darwin Knows Exists
This is why we’re building Darwin Knows.
Not to build the 101st chatbot. Not to surface the 10th best ESG screen. But to rethink the whole model.
We are building:
Pipelines of intelligence, not isolated workflows
Agentic personas that specialise, learn, and evolve
Contextual layers of knowledge that plug into portfolios, strategies, alerts, and decisions
Access models that empower both professionals and everyday investors to engage with financial markets on equal footing
We are building an AI-native distribution model for financial intelligence.
And just as DMA compressed commissions, we believe AI agents will compress the cost of knowing.
We’ve Seen This Play Before
The signs are all around us.
Commissions went from 90bps to 7bps.
Teams of 150 became teams of 15.
“High touch” gave way to “low latency.”
Entire business models vanished, rebuilt in code.
What happened to execution is now coming for intelligence.
It may not be tomorrow. But it is close. The convergence of cloud, compute, data, and agents ensures it.
And when it does, the entire value chain of research, advisory, planning, and insight will be rebuilt.
The knowable will be made available.
The accessible will become affordable.
And the intelligent will become participatory.
We do not fear this change. We welcome it.
Because the future of intelligence is not gated.
It is open.
It is agentic.
It is Darwinian.
Final Thought: Be Certain
I’ve said this before, and I’ll say it again.
It may not be tomorrow. But be absolutely certain. It is coming.
Just as electronic execution reshaped capital markets forever, so too will AI agents reshape the landscape of financial intelligence.
The knowledge graph is being built.
The signals and patterns are forming.
The next wave has already begun.
Daryl Bowden, Founder
Darwin Knows.