OneTrack started in 2017 with a different name and a different form factor.
At the time, Marc Gyöngyösi was building autonomous drones for cycle counts and inventory checks in warehouses. The company is still legally Intelligent Flying Machines, doing business as OneTrack.
The product changed. The problem did not.
Where the Problem Started
The origin goes back to Marc's work on robotics projects connected to BMW in Munich. Manufacturing robotics was tuned for precision and control loops. Warehouse operations next door were operating with much less visibility.
That contrast shaped the thesis that still drives the platform:
- Manufacturing systems had strong control loops.
- Warehouse operations often depended on partial records and manual checks.
- Physical operations needed better ground truth, not just better reports.
The First Product: Autonomous Drones
Early traction came quickly. The team demonstrated at TechCrunch Disrupt and ran pilots with customers willing to test new approaches.
The technology worked, but the operating model had constraints:
- Safety risk in live environments with spinning blades near people and equipment
- Operational complexity for everyday deployment at scale
- Harder path to durable product-market fit despite technical feasibility
As Marc put it on Code Story: "The problem was right. The form factor was wrong."
The Pivot: From Flying Robots to Ground Truth
The team pivoted from drones to camera systems on material-handling equipment already moving through the building every day.
Instead of sending a flying platform through a facility, OneTrack focused on capturing what actually happens in routine operations and converting that signal into usable decisions.
That shift changed the deployment model:
- Better safety and fit for live operations
- Faster and more repeatable installs
- A stronger data foundation tied to observed behavior, not only transaction timestamps
Building for Scale
OneTrack now processes extremely large data volumes, with billions of images analyzed daily across customer environments.
A core challenge has been building a stack that compresses raw visual signal into useful operational outcomes:
- Billions of frames
- Thousands to tens of thousands of candidate events
- Verified incidents and behavior patterns that teams can act on immediately
This is where architecture mattered most. In the episode, Marc describes many MVPs across the journey, not one:
- Early drone autonomy demos in a 10,000 sq ft warehouse in Chicago
- First multi-sensor forklift deployments that had to work every day
- First operator-facing product surfaces beyond raw database outputs
- First agent workflows, initially launched as a small chat interface beside safety events
Each step required hardening the platform without losing iteration speed.
Why the Team Focuses on Iteration
A recurring theme in the conversation is simple: there is no single "MVP moment." There are many MVPs that compound.
The operating principle is to stack small wins while keeping the technical foundation modular enough to adapt quickly as model capability changes.
That matters in AI because planning on multi-year static roadmaps is less useful when the underlying model layer changes in months.
The Agentic Shift
Over the last year, OneTrack expanded from analytics and visibility into agentic workflows.
What began as a small chat interface next to event workflows became broader agentic architecture:
- Agents that reason across operational context and connected systems
- Agents that investigate exceptions and draft next actions
- Humans focused on policy, approvals, and edge cases requiring judgment
In Marc's words, the goal is an "application built by agents, used by humans."
Long-Term View
Another theme in the episode is company design, not just product design.
OneTrack has taken a long-horizon approach to customer partnerships and platform development. The focus is durable operating value in real facilities, not short-term theater.
The broader thesis remains the same: enterprise software for physical operations should understand what is actually happening on the floor, then help teams act quickly and consistently.
The full Code Story episode covers the early drone years, the pivot, and how OneTrack is building agentic systems for physical operations.
Listen to the episode →
Want to go deeper on the platform direction discussed in this story?
- Explore AiOn - Agentic AI platform for warehouse and logistics operations
- Read the Agentic AI guide - Definitions, architecture, and production use
- See Supply Chain use cases - Where agents deliver immediate ROI
Related Articles
- How We Built an AI That Understands Warehouses - The technical foundation behind OneTrack's computer vision stack
- What is Agentic AI? - A practical framework for agentic systems
- The AI Adoption Gap Is the Real Opportunity - Why execution matters more than tooling
- Agentic AI for Supply Chain - How agents are transforming logistics operations