Write Non-Negotiables AI Can Actually Check
A flag on a submitted asset has three possible causes, and the requirement you wrote is the one nobody checks first. The test for whether a non-negotiable can actually hold.
A flag on a submitted asset has three possible causes, and only one of them is the creator. The second is the system. The third is the requirement you wrote, and in our experience it is the one nobody checks first.
That third category is worth your attention because it is the only one you control directly, and because it fails in a direction most teams do not expect. A badly written non-negotiable does not usually produce a wrong flag. It produces a clean pass on a requirement that was never evaluated at all.
The short version: AI Asset Analysis checks each submitted asset against the requirements you set in your Brief. It reads one file at a time. Any requirement that cannot be answered by looking at one file, on its own, with no other context, is a requirement it cannot hold for you.
TL;DR
- 🎯 One asset, one look. Each submission is scanned on its own. A requirement spanning three deliverables cannot be checked against any one of them.
- 🤫 Unverifiable requirements fail quietly. They do not throw a flag you can investigate. They pass, and you read the pass as confirmation.
- ✍️ The stranger test. If someone with no knowledge of your brand could not give a yes or no by looking at the file alone, it is not a non-negotiable yet.
- 📋 Nothing gets deleted. Requirements that fail the test move to Concept, Creative Direction, Style Guide, or Timeline, where they steer the creator instead of gating the asset.
- 🔁 Check before launch, not after. Brief Analysis reviews the Brief before it goes out. Fixing a requirement pre-launch costs one edit; fixing it post-launch costs a reshoot cycle across every creator in the Brief.
What does AI Asset Analysis actually evaluate?
It evaluates what is observable in a single submitted file, measured against the requirements written in that Brief. Nothing more.
When a creator uploads, the scan runs automatically against the Brief's requirements and returns structured feedback, typically inside an hour. A passing asset gets a badge. A flagged asset generates a notification and an automated chat message that both the creator and the brand can see. Creators are told to wait for your instruction before editing, so all feedback arrives in one round rather than three.
That workflow is why requirement quality compounds. Every requirement you write gets applied to every asset in the Brief, consistently, without fatigue. Written well, that is the whole point. Written badly, you have automated a bad question.
The constraint that matters: the scan sees the asset. It does not see the platform post, the calendar, the contract, the creator's other deliverables, or your internal brand debate.
Why do vague non-negotiables cause more damage than wrong ones?
Because a wrong flag announces itself and a vague requirement does not.
If a requirement produces a flag you disagree with, you see it, you dismiss it, and you may send a change request. The system told you something. Annoying, but visible. Cohley grays out the flag reason once dismissed, so the record of your decision stays on the asset.
A requirement like "make it feel premium" behaves differently. There is no observable state that satisfies or violates it. So it does not reliably generate a flag, and the asset moves through looking clean. You approve on the strength of a check that never happened, and you find out at activation, when the asset is already in a paid rotation.
This is the failure mode we care most about. A false flag costs you five minutes of irritation. A silent pass costs you a campaign asset you thought was cleared.
The pattern is not specific to creator content. As HappyFunCorp's Jonathan Zaleski put it in InfoWorld, teams working with agentic systems have to separate the requirements best enforced by machines from those that still need human judgment. Objective and measurable criteria scale. The emergent and intuitive ones do not, and pretending otherwise just moves the failure later in the process.
What is the test for a good non-negotiable?
Could a stranger with no context give you a yes or no by looking at this one file, and nothing else?
If yes, it is a non-negotiable and it belongs in the non-negotiables section. If no, it is direction, and it belongs somewhere else in the Brief. Both are legitimate. Only one of them can be enforced at the asset level.
Run the test out loud on your last Brief. Most teams find two or three requirements that do not survive it.
The five categories that reliably fail
Each of these reads like a rule but cannot be settled by looking at one file.
- Cross-asset"Vary the setting across your three videos."Each asset is scanned alone. No single file contains the comparison.
- Caption and post"Tag @brand and use #brandpartner."The caption lives on the published platform post, not in the uploaded file.
- Process, timing, history"Film within one week of receiving the product." "Don't reuse a setup from a past Brief."Neither is visible in the asset. Timing and history are records, not pixels.
- Rights and licensing"Only use music you have cleared."A rights determination is a legal judgment about permission, not an observation about a file.
- Vague or subjective"Make it feel premium." "Keep it authentic."No pass state exists, so no fail state exists either.
The rights row deserves a specific caution. Automated checking can surface signals that warrant a human look. It does not constitute legal verification, licensing, or clearance, and no requirement you write should be read by your team as producing that. Keep rights language in your usage terms, and point creators to the licensed library rather than asking them to self-certify a legal conclusion in a non-negotiable field.
How do you rewrite a requirement that fails the test?
You split it. The intent stays in the Brief and moves to the section built for it. The enforceable residue stays as the non-negotiable.
Where each intent belongs
Nothing is deleted. The intent moves to the section that can carry it.
- "Make it feel premium."Creative Direction, plus reference images in Style Guide"Shot on a plain, uncluttered surface in natural light."
- "Show the product in use."Concept"Product visible in hand and actively being applied for at least three seconds."
- "Tag @brand and use #brandpartner."Deliverables, verified at the postRemoved from asset-level requirements
- "Vary the setting across your three videos."ConceptMade per-deliverable: "Deliverable 2 is shot outdoors."
- "Film within one week of delivery."TimelineRemoved from asset-level requirements
- "No competitor products."Stays"No competitor branding or packaging visible in frame."
Notice what happened to the last two rows. "No competitor products" was already close, and only needed the ambiguity resolved: does a competitor's unbranded jar count? Say which you mean. "Film within one week" was never a content requirement at all. It was a deadline wearing the wrong hat.
The Brief wizard already separates these sections for a reason. As we put it in the seeding fit test, non-negotiables are rules and they should read like rules. Everything that is not a rule is direction, and direction works better when it is not pretending to be a gate.
Who should own this, and when?
- If you are a brand operator writing Briefs: run the stranger test on your non-negotiables before launch, and use Brief Analysis on the full Brief. Your target is zero requirements in the five failing categories.
- If you are a CSM reviewing a customer's Brief: the fastest diagnostic is to read only the non-negotiables list, with no other context, and mark every line you could not adjudicate from a single file. That list is the conversation.
- If you are auditing flags after the fact: sort by cause before you sort by creator. Separate the flags caused by the requirement from the flags caused by the content. Acting on the second category while the first is unresolved produces change requests creators cannot act on.
- If your team is debating a subjective standard internally: resolve it before it reaches a Brief. A requirement your own team disagrees about will not be settled by an automated check.
What this does not change
Automated checking is a filter in front of your queue, not a replacement for your judgment. Cohley's public position is that AI Asset Analysis checks every submission against your Brief's non-negotiables before it reaches your queue. That is a bounded claim on purpose. It does not certify compliance, and it does not settle brand-fit questions that require a person who knows your brand.
The value is narrower and more durable than a compliance guarantee: consistent application of the criteria you were specific enough to write down. Which is why the writing is where the work is, not the model.
Frequently asked questions
What is a content non-negotiable?
A non-negotiable is a specific rule a submitted asset must meet, set by the brand in the Cohley Brief. It differs from creative direction because it is binary: the asset either satisfies it or it does not. Non-negotiables are the criteria AI Asset Analysis checks each submission against before the asset reaches the brand's review queue.
Why did AI Asset Analysis flag an asset that looks fine?
A flag has three possible causes: the requirement was written in a way that cannot be evaluated reliably, the system needs tuning for that requirement type, or the content did in fact miss the mark. Check the requirement wording first. If the requirement is ambiguous, the flag is a symptom rather than a finding.
Can AI Asset Analysis check captions and hashtags?
No. The scan reads the uploaded asset, and captions live on the published platform post rather than in the file. Tagging and hashtag requirements belong in your deliverables and should be verified at the post. Writing them as asset-level non-negotiables produces no reliable check.
Does a passing scan mean the asset met all my requirements?
It means the requirements that could be evaluated from that file were evaluated. Requirements in the five failing categories, including cross-asset, caption, timing, rights, and subjective standards, do not produce a dependable result. A pass on those is not confirmation.
What happens when I dismiss a flag?
The flag reason is grayed out on the asset, preserving the record that a person reviewed and overruled it. Creators are instructed not to edit until they hear from you, so dismissing a flag without messaging the creator leaves them waiting rather than reshooting.
Should non-negotiables be strict or loose?
Strict and specific, but only about things observable in the asset. Looseness causes silent passes; strictness about unobservable things causes creator confusion and unactionable change requests. Precision on the observable is the goal, not severity.
Where do requirements go if they fail the stranger test?
Concept, Creative Direction, Style Guide, Deliverables, or Timeline, depending on intent. Nothing is deleted. The intent stays in the Brief where creators will read it, and only the enforceable portion remains as a non-negotiable that gates the asset.
Related posts
- Aug 13, 202612 min read
Product Seeding Works Best When the Product Is the Pitch
In seeding, the product is the creator's compensation. Here's a five-part test for whether yours can carry that weight.
Zach Chmael - Aug 31, 20266 min read
The Quality vs. Volume Problem in UGC, and Why Brands Keep Choosing Wrong
Brands keep buying UGC based on creator count and getting burned by unusable content. How much a brand can publish depends on how each submission is briefed, reviewed, cleared, and activated.
Erik Graber - Sep 4, 20269 min read
Have You Outgrown Your UGC Video Tool? A Fit Test for Brands Scaling Creator Content
Most brands start with a self-service UGC video tool for good reasons. Five signals the job has changed from sourcing videos to running a content operation, plus three reasons to stay exactly where you are.
Michelle Green
Next step
Open your most recent Brief and read only the non-negotiables list. Mark every line you could not adjudicate from a single file with no other context.
Then run the Brief through Brief Analysis before the next launch.