Your ask has a ceiling, and it usually isn't the band
You are told two numbers before you negotiate: the band (what the role is posted at) and your target. Most advice assumes the band is the wall, and the job is to get as close to the top of it as you can.
That is often the wrong wall.
Two different ceilings
There are usually two limits in the room, and they are not the same number:
- The band ceiling. What the role is budgeted for. This is what you can see.
- The authority limit. The highest number the person across the table can approve without asking anyone. This is what you usually cannot see.
If the authority limit sits below the band ceiling, then the band is not your constraint — the approval path is. And the two produce completely different advice:
| If the binding constraint is… | The right move is… |
|---|---|
| the band ceiling | build the case for sitting at the top of the band |
| the authority limit | find out whether an exception path exists, and who has to be convinced |
| the band ceiling and no exception path | stop pushing on base, and move to non-cash terms |
Asking harder for base salary when the real problem is an approval limit does not just fail. It spends your credibility on the one thing the other person cannot give you.
Why the model treats this as a constraint, not a vibe
This is the part that a language model reading your email cannot do reliably, and it is why the engine is arithmetic rather than persuasion.
Your situation becomes an explicit model: the band, the approval limit, whether an exception path exists, who has to sign off, and how much urgency is on their side. The engine then searches the space of situations consistent with those constraints and reports which actions survive — with the reason codes that produced them.
So when a report says “get to the approval path before you talk about the number again”, that is not a hunch. It is what follows from the constraint you gave us, and you can see the reasoning.
What this means for your ask
Three practical consequences:
- Find the approval limit before you optimise the number. A question like “if we agreed on a figure today, is that something you can confirm, or does it need to go somewhere else?” costs nothing and tells you which wall you are facing.
- Do not read a refusal as a ceiling. “That’s more than I can do” is information about authority, not about budget. Those two require opposite next moves.
- If the base is capped, change the currency. Sign-on, a review date in writing, an early band re-evaluation, or a title change are all different constraints — and one of them may be inside the authority the person actually has.
The honest limit of this
None of the above tells you what will happen. If you do not know whether an exception path exists at this company, the model will say so, and the report will tell you that resolving that is your next move rather than inventing an answer for you.
That is the trade: you get a specific next step and the reasoning behind it, and you do not get a prediction. Anyone offering you the prediction is guessing about a company they have never seen.