Where to Draw the Line
Vishal Sachar
Co-Founder & CEO of CLRT
The teams that ship reliable AI and the teams that ship liabilities are not divided by whether they use AI. They both do. They are divided by how their outputs get verified. That single variable, the rigor of the check, is the whole game.
There are three broad modes, and they sit on one line. At the loose end is vibe coding: casual prompts, a quick glance to see if it seems to work, disposable output. In the middle is structured assistance: detailed prompts with real constraints, manual testing and spot checks, genuine features built into real systems. At the tight end is agentic engineering: formal specifications, automated evaluations, deployment gates, production systems running at scale. The line runs from speed at one end to reliability at the other.
The common mistake is to read this as a moral hierarchy, with vibe coding as the sin and agentic engineering as the virtue. It is not. It is a fit decision. A throwaway prototype you will delete on Friday deserves the speed of vibe coding, and wrapping it in formal evals would be waste. A payment system deserves the full rigor, and shipping it on vibes is how you end up explaining a breach.
The skill, the actual professional judgment, is drawing the line in the right place for each task. Not defaulting to one mode for everything, but matching the rigor to the stakes, deliberately, every time.
The differentiator is not whether you use AI. It is how the output gets checked. Draw the line by the stakes, every time.
A deeper dive
Verification means something concrete and different at each tier, and naming it removes the mystique. At the vibe tier, verification is a human eyeballing the result, fine when being wrong costs nothing. At the structured tier, it is a person testing the output against expectations and spot-checking the edges, with a human approving each meaningful change. At the agentic tier, the specification itself becomes the test: automated evaluations run on every change, deployment gates block anything that fails, and the system corrects itself against those checks before a human ever sees it. The reason the correct line tracks the stakes is the cost of an uncaught error. Reversible and cheap tolerates a glance. Irreversible and expensive demands automated gates and a checker independent of the maker. Most real work lives in the structured middle, and the most common failure is not picking the wrong tier once, it is unconsciously applying one tier to everything, moving too slowly on what is trivial and too fast on what is dangerous.
The sequence
- 01
Vibe coding
Casual prompts, a quick glance, disposable output.
- 02
Structured assistance
Detailed prompts with real constraints, manual testing and spot checks.
- 03
Agentic engineering
Formal specifications, automated evaluations, and deployment gates.
Work with CLRT
Knowing how much rigor each workflow deserves is a strategy decision, not a technical one. CLRT helps you set that line across your business, so you move fast where it is safe and slow where it counts. Start with a conversation.
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.
