The single highest-leverage use of AI in brand work isn't generating positioning. It's pressure-testing positioning you already have. Most teams skip this step because it's the unglamorous half of the work. They build the positioning, present it to leadership, get sign-off, ship it. Then they find out in market that it doesn't hold up.
The AI version of pressure-testing takes 20 minutes. It catches most of what would otherwise blow up later. It's the cheapest insurance you can buy on a strategic call.
HOW TO ACTUALLY DO IT
Paste your draft positioning into a chat. Then run it through five attacks in sequence. The attacks are designed to catch specific failure modes. Each one takes a single prompt.
Attack 1: The competitor pitch. "Here's our positioning: [paste]. Now imagine a sharp competitor in our category. Write the slide they would use to pitch against this position to a customer who's deciding between us. Be ruthless. What's the strongest version of their counter-argument?"
This catches positions that have an obvious counter. If the competitor's pitch writes itself, your position has a structural weakness.
Attack 2: The dismissal test. "An indifferent customer reads our positioning and dismisses us in three seconds. What was the line that made them dismiss us? What was the unstated assumption we made that made it easy to dismiss?"
This catches positions that sound powerful internally and read as nothing externally. The dismissal usually targets the implicit assumption that the team forgot to interrogate.
Attack 3: The assumption stack. "Identify every assumption this positioning rests on. List them in order from most-load-bearing to least. For the top three, tell me what would have to be true for each, and how confident we should be that it actually is."
This catches positions that are conditional on something the team hasn't actually verified. Sometimes the position is right but rests on an assumption that's about to break.
Attack 4: The opposite play. "Write the strategy memo for the exact opposite positioning. Not 'different.' Opposite. What would have to be true for that to be the right call? Is any of it true?"
This catches positions that look right by default. If the opposite position is more defensible than yours, you've made the wrong call.
Attack 5: The eulogy. "If our brand disappeared tomorrow, what specifically would the world be missing under this positioning? Be specific enough that the answer couldn't be true of any competitor."
This catches positions that sound like everyone else's. If the eulogy is generic, the position is generic.
WHAT TO DO WITH THE RESULTS
Most positions will fail one or two of these attacks. That's normal. The question isn't whether they fail. It's whether the failures are fixable through sharpening, or whether they reveal that the position itself is wrong.
Sharpenable failures: the language is too vague, the specific differentiator isn't named, the customer is too broadly described. Fix the language and re-run.
Structural failures: the competitor pitch writes itself, the assumption stack rests on something that isn't true, the eulogy reads as generic. The position needs to be different, not just better-worded.
WHY THIS WORKS
The AI is bad at generating original strategy. It's good at finding flaws. Generation requires conviction. Critique requires pattern matching. The model is far more capable at the second than the first.
Most teams use AI backwards: they ask it to generate (where it's weak) instead of asking it to critique (where it's strong). Flip the workflow. Generate yourself. Critique with AI. The output is a position that's been stress-tested against a thousand pages of competitor pitches and customer dismissals before it ever leaves the room.