Contents
Rationalizing the AI transition
It’s completely obvious that AI has already dramatically changed industries and it is still largely unknown what all the impacts will be.Through my experience building an AI orchestration platform, I have realized that deploying commercial AI tools is merely a baseline prerequisite. True competitive advantage is never built by renting the exact same off-the-shelf tools as your rivals.
AI as a competitive advantage
Simply using new AI tools (Claude Code, Gemini, etc) is a pre-requisite to even begin this exercise. There is likely nothing to be gained from this perspective if AI tools are not being deployed, scaled, and used by a company. But turning a transformative technology like AI into a competitive advantage is a lot more than just using tools. True competitive advantage is not built by deploying the same commercial tools as your rivals.
Although I wasn’t alive or participating in industry when the Internet was developed, the Internet transition provides the exact transformation blueprint for the AI era. Companies that just bought any ISP subscription survived, but the companies that re-architected their transaction pipelines around the web became titans. The exact same rule applies to AI. Buying generic seats to a third-party tool is just buying an ISP subscription. Building a custom, vertically integrated agent harness is re-architecting the business.
The Product Factory in action
I like the abstraction of thinking of an organization as a product factory. The goal is to eliminate every area of friction in the factory to maximize the velocity and quality with which products are brought to customers. To quickly build a competitive advantage, organizations should consider their tools, processes, workflows that bring product to customers as a unified platform. This now allows for AI to be applied strategically to every facet of the business. With the unprecedented speed of software development and deployment, custom software and tools should be built to ruthlessly eliminate friction in organizations’ product factory. This level of vertical integration builds a highly defensible competitive advantage that can be tuned and scaled as AI continues to evolve.
By deploying a custom agent harness integrated directly into our tools and infrastructure, we achieved a 10x - 15x increase in developer velocity compared to team members using generic agentic IDEs (Windsurf, Claude Code, etc).
1. The Control Planes
We didn’t just stop at code generation. We integrated our agent harness directly into our version control and build infrastructure, making pull requests and CI/CD pipelines an active control surface. Instead of a developer context-switching to fix merge conflicts or fix a formatting suggestion, background agents interpret the PR commentary and execute branch fixes autonomously.
2. Giving our Agents Eyes
Our organization builds and sells PCs and peripherals, so we hooked up cameras and made those native tool calls for our agent harness. Our background agents could now drive and validate its changes in the physical world. This unlocked long-running autonomous loops where the agent could implement complex lighting effects and validate its changes in real time.
3. Observability
Every LLM trace, every tool call, every decision our agents make is fully visible. This gives us tremendous control over how our agents operate, perform, and generate value for our organization. Every knob helps us tune the platform to be maximally valuable to our organizations’ needs. We also can proactively detect platform degradation for any area of our business. Maybe an AI lab degraded their model to prepare for the next training run, maybe a team has shifted their tech stack - whatever it is the platform can detect and alert us to take the necessary steps before the business is impacted. As companies start to reconcile their AI costs with value created, our platform allows us to map every single LLM call to the value it created.
Expanding the Moat
We’re now extending our platform into the rest of the Product Factory.
Product roadmaps: Augment our factory with competitive analysis tooling that gathers data on how our products are performing in the field, monitor social media and customer feedback and use that to shape roadmaps.
Hardware development: Build debugging harnesses that connect our firmware development agents to the physical world, accelerating new development and root-cause analysis, debugging, or validation.
Operational integrations: Connect pricing, sales, and support into the platform so all product decisions are rooted in rich, unified data that synergizes.
By leveraging AI in this way, organizations’ data becomes a valuable resource that unlocks further tuning and visibility into how work gets done. Inference can be routed, degradation can be detected, emergent optimizations can be identified. By abstracting to the concept of a Product Factory, we can expose as many knobs as possible to deepen the moat and expand the competitive advantage.
In Closing
Leveraging AI as a transformative technology means rethinking how businesses operate and remain competitive. To build a durable moat, it is essential to identify as many areas within an organization to optimize and vertically integrate. To see more details of the platform I built, read the rest of my writeup: AI Orchestration Platform.