From Apps to Agentic Workflows
0x41434f
This timeline illustrates the evolution of enterprise workflows, from manual processes to agentic AI.
This post began as a way to make sense of a growing obsession. Over the past year, I’ve been circling around one idea again and again: how work actually gets done, and how much of it still depends on brittle systems, human handoffs, and silent inefficiencies. It started with reading, then turned into conversations, fieldwork, side projects, and late-night sketches. I worked undercover at fulfillment and warehouse centers, observed the small failures that snowball into lost revenue, and built internal tools for startups dealing with security and compliance. I sifted through acquisition data, watched the chaos of onboarding unfold at scale, and studied how organizations adopt or resist change.
Lately, my work has turned toward something harder to name. I’ve been building in stealth on a new kind of product, shaped by everything I’ve seen. It’s not just about better apps. It’s about workflows that think for themselves. The way I describe it to friends is this: imagine if the systems you used at work could not only understand your intent but act on it, adapt to new constraints, and collaborate with other systems without waiting for you to click a button. That is the shift I’m exploring. Away from tools that sit passively in the cloud, toward intelligent agents that operate across departments, learn from interactions, and improve as they go.
This isn’t just a theoretical idea. We’ve spent decades optimizing productivity software, but in many industries, the actual work remains disjointed. IT tickets still get lost. HR teams still juggle spreadsheets. Compliance still means chasing documentation across five different tools. Most software isn’t broken, but it is bounded. It does exactly what you tell it to and nothing more.
What excites me about this new generation of agentic systems is that they start where traditional software stops. They don’t wait for humans to orchestrate every task. They observe, learn, infer, and take initiative. In cybersecurity, this can mean real-time response to threats. In operations, it might mean spotting bottlenecks in shift coordination before they escalate. In compliance, it is the ability to enforce rules not just at the point of audit but continuously, across systems and roles.
The stealth project I’m working on now is shaped by these patterns. It draws on everything from my fieldwork in logistics to the policy research I’ve done around data privacy. It borrows from the logic of agents, the structure of enterprise workflows, and the human desire to spend less time managing software and more time doing meaningful work. It isn’t about replacing people. It is about giving them systems that act like collaborators instead of dashboards.
As I keep learning, I’m also mapping out where this leads. I’m interested in agent orchestration, multi-agent collaboration, reasoning chains, and the practical challenges of grounding AI in business logic. I think we are just at the beginning of building tools that understand not just language, but work itself. And I’m convinced that somewhere between logic, law, and machine learning lies the foundation for a better future of work.
This project is one attempt at building that.