The Model Is the Engine. The System Is the Car, the Road, and the Traffic Laws.
Vishal Sachar
Co-Founder & CEO of CLRT
Almost everyone watching AI is watching the wrong thing. They track which model is smartest this month, which one topped the leaderboard, which lab shipped what. It is the most visible part of the field and the least decisive. Swapping engines does not build you a better car.
Most of what determines whether an AI agent actually works is not the model at all. It is the system built around it: the instructions that give it a role, the tools it can reach, the memory it reads and writes, the guardrails that stop it crossing lines, the orchestration that sequences its steps, and the verification that checks its output. As a rough split, call it nine parts system, one part model. The engine matters, but it is the smallest part of the machine.
This is the best news a business could ask for. You do not win by having secret access to a smarter model, because everyone can rent the same ones. You win by building the better system around it, and that system is something you can own, improve, and defend. The advantage moved from what you can buy to what you can build.
The image to hold is mechanical. A Formula One engine bolted into a shopping trolley loses a race to a family car on a proper road, because once the engine is good enough, the car, the road, and the traffic laws decide the outcome. The model is the engine. Everything that makes it useful, safe, and repeatable is the rest of the vehicle.
The model is the engine. The system is the car, the road, and the traffic laws. Stop shopping for engines and start building the car.
A deeper dive
It helps to see the system as layers wrapped around the model. Closest in sits the framework layer: the instructions and rule files, the tools and connectors the model can call, the orchestration logic that decides what happens in what order, and the guardrails and hooks that constrain it. Around that sits an operational layer: the session and memory store that gives the agent state, the evaluation and testing that tells you whether it is actually working, and the observability that lets you see what it did and why. Around that sits deployment and scaling. The model is one component in the centre of all of it. The reason the ninety-ten framing matters is not the exact ratio, it is the implication: every layer in that diagram is something you control and tune independently of which model sits at the core. That is also what makes a system engine-agnostic, able to take a better or cheaper model later without rebuilding everything around it.
Work with CLRT
The model is the part you rent. The system is the part that wins. Building that system around your highest-value workflows is the work CLRT does. Start with a conversation about where yours is weakest.
Vishal Sachar
Vishal Sachar is the Co-Founder and CEO of CLRT, where he helps UAE businesses make sense of applied agentic AI and put it to work. He writes on agentic systems, AI governance, and the economics of automation. Reach him at vishal@clrtstudio.com or on LinkedIn.
