What is an AI dark factory?

A dark factory is a plant that runs with nobody on the floor, so the lights can stay off. In software, an AI dark factory is a delivery pipeline where AI agents do the routine work end to end: they pick up a ticket, reproduce the problem, write the fix, run the tests and open a pull request. People decide what matters, review the changes and talk to customers.

The loop

1. 1Intake. A problem is reported and becomes a ticket in the tracker.
2. 2Reproduce. An agent reads the ticket and recreates the problem on the same page.
3. 3Fix and verify. It changes the code and runs the tests.
4. 4Review and ship. It opens a pull request; a person reviews, merges and ships.

Why the ticket decides

An agent cannot walk over and ask what the reporter meant. Everything it knows comes from the ticket, so a vague one (“the button doesn’t work”) stalls the whole line. A good ticket carries what the person saw, where it happened, in which browser and screen size, and what the console said at that moment.

Where Prezenca Critique fits

Prezenca Critique is the intake. Anyone on your live site marks up a screenshot and writes a sentence; the ticket lands in Jira or your inbox with the annotated screenshot, the page URL, the browser, operating system and screen, and the console log with JavaScript errors. That is the context an agent needs to act without asking back.

We do the reporting and the ticket. The agent is yours: any agent that works from your tracker can take it from there.

What stays human

Priorities, product decisions, the review before anything ships, and the conversation with the person who reported the problem. The factory runs the routine; people run the factory.