What Makes an AI Brief Usable by Someone Who Did Not Write It?
According to Dr. Jonah Tebaa, an AI brief becomes usable by others when it passes the Cold-Run Test, meaning a colleague runs it without guidance to achieve comparable results. To survive handovers, an inheritable brief must explicitly document five elements: why it exists, excluded materials, a named owner with a last-checked date, the expected output format, and when not to use it. Furthermore, any new reader, material, or quality standard requires splitting it into a separate brief.
Many small firms have a person who is "good at the AI tool". Her colleagues use the same software and get weaker work, and the gap is usually put down to skill. Dr. Jonah Tebaa reads it differently. In his view, the tool is rarely the variable. The variable is a brief that exists only in one person's head, and a brief like that cannot be handed over.
A saved line is not a brief
Dr. Tebaa opens with a scene. A manager is on leave and the monthly client summary is due. A colleague finds a saved instruction reading "make it sound like us", runs it, and receives something unusable. The manager's own results are good because she silently supplies the missing context each time: which client wants brevity, which sheet holds the signed-off figures, which promises the summary must never make.
His conclusion is that an instruction only its author can run well is a personal habit, not a firm capability. He distinguishes it from the short specification he has described before, which serves a single task and ends with it. A brief a team reuses every month, he argues, is a work document and needs to survive a change of hands.
The Cold-Run Test
The framework he offers is a test rather than a template. A colleague who did not write the brief runs it once, with no explanation. Whatever goes wrong is noted against the part of the brief that should have prevented it. Nobody repairs the output by hand; the team changes only the brief and runs it again. The brief passes when two different people get comparable results.
His reasoning is that an author cannot test their own brief, because they cannot unlearn what they know. Watching a capable colleague stumble reveals which assumptions were never written down, and a quiet working day is a cheaper time to learn that than the day the author is away. Rescuing the output, he adds, hides the gap from the next person.
What a brief must carry to be inherited
Dr. Tebaa says a brief meant for reuse has to hold more than a one-off specification does, and he names five things.
- Why it exists: the recurring decision or reader it serves.
- What stays out, and why: materials deliberately excluded, with a reason, so no one adds them back.
- A named owner and a last-checked date: one person who may edit it, and proof that it still works.
- What it must hand back: the shape of the output, plus a list of what the tool could not find.
- When not to use it: the cases it was never built for.
His worked comparison uses a generic monthly summary. The weak version is one sentence. The inheritable version is not much longer, yet it names a reader, the approved figures, a ban on stating delivery dates, a fixed layout, an owner and a case in which to stop and ask. A stranger could run it.
Splitting and retiring briefs
Successful briefs, Dr. Tebaa notes, attract extra duties. A summary instruction becomes a summary, a covering email and a board version, and its rules begin to conflict. His rule is that a different reader, different materials or a different test for good means a different brief, with its own owner. He also asks firms to review each brief on a set date and again whenever the underlying software changes, and to retire instructions nobody uses, because an old saved brief looks authoritative to the next person who finds it.
What the idea asks of an owner
The practical demand is modest. Dr. Tebaa suggests choosing one repeated task, writing down why it exists, what stays out, who owns it, what it returns and when not to use it, and then having a colleague run it cold. It takes an hour or two, and the firm ends up with a piece of working method that survives holidays, resignations and software updates. For owners in Lebanon, the wider region and elsewhere who rely on one or two enthusiasts, the idea turns a private skill into something the whole team can use.
The full argument, with the worked illustration and the practical steps, is available in his own words on jonahtebaa.com.