What Dragonflies Taught Me About Agentic Systems in Finance

Originally posted on LinkedIn · April 16, 2026

I spent part of this weekend watching dragonflies hunt in a wetland near my home.

~95% prey interception rate.

In financial forecasting or real-time decision systems, that level of consistency is extremely difficult to replicate.

But the insight isn't capability. It's structure.

Staying in the loop

Dragonflies operate through continuous perception-to-action loops, adjusting in real time. They don't predict and then act. They stay in the loop.

That got me thinking about agentic systems in finance.

Most enterprises already have layers of automation: spreadsheets, workflows, analytics platforms. Each works well in isolation. But collectively, they fragment data, logic, and execution.

The issue isn't a lack of models. It's a lack of environmental coherence.

Even strong systems degrade in noisy, inconsistent environments.

The constraint is the feature

Another pattern stood out. Dragonflies operate within strict boundaries. That constraint is what makes them reliable.

The same applies to agentic systems: clear scope, defined interfaces, and guardrails improve outcomes more than unconstrained capability.

Selectivity over volume

Dragonflies act when signal quality is high. They hold when it isn't.

In financial systems, unnecessary action has real cost.

Where performance compounds

The shift underway is not just toward better models. It's toward better systems:

• cleaner inputs
• tighter feedback loops
• stable environments with constrained execution

Less focus on which model wins. More focus on what surrounds it.

That's where performance compounds.