“I’ll get back to you.” Nothing in that sentence is untrue. They will raise it, or intend to. What it omits is the only thing I needed: that the decision was never theirs, that nothing I said in that meeting could have moved it, and that I had spent three weeks building a case for someone who holds no right to answer it. I have asked for more people maybe a dozen times, across multiple companies and many managers. Not once did anyone open with this isn’t mine to decide.
So I kept improving the argument. Better numbers, tighter framing, a cleaner link to revenue at risk. All of it aimed at a node that was never going to render a verdict, because its function in that conversation was transport, not judgment. The sentence has the grammar of a commitment and the mechanics of a referral, and it lets everyone in the room believe a case is being evaluated when it is being carried elsewhere. Because it sounds like advocacy, questioning it makes you look ungrateful.
Let me be precise about what I am not saying. Nobody in these conversations pretended to be the decider. They simply didn’t mention that they weren’t, and I think that omission is closer to structural than personal. “This isn’t mine to decide” is a hard sentence to say out loud in a role whose entire standing rests on owning the team’s outcomes. Nothing in the title, the org chart, or the meeting invitation obliges anyone to say it. So it goes unsaid, reliably, across companies, by people who meant me well. The silence is not a character trait. It is what the role costs.
The constraint is real
A team lead can’t approve headcount. Neither can a UX director in most organizations. That isn’t an accident. Headcount is the most irreversible commitment an organization makes. It compounds, it survives the project that justified it, and it is nearly impossible to unwind without damage. In a regulated, audited environment it is also one of the few numbers that has to reconcile cleanly to a plan someone signed. Concentrating that decision high up is a defensible design choice.
Fine. But if the answer to “give me more capacity” is legitimately upstairs, then the other lever should be local. Fewer commitments. Cut the scope. Tell the stakeholder that the thing they asked for in March isn’t happening this year. That decision costs nothing in budget terms and is fully reversible, and it is the obvious counterpart to a centralized hiring process.
It is also, in my experience, escalated. Every time.
The asymmetry that actually matters
What I expected to find was a story about managers who avoid uncomfortable local decisions. What I actually found was a layer with full authority over the shape of the work and none whatsoever over its size. Sequencing, assignment, technical approach, who pairs with whom, what quality bar we hold, which candidate gets the offer for an already-approved slot: all local, all decided, often decided well. Anything that changes the total volume of commitment, in either direction, leaves the room. More people: upstairs. Fewer promises: upstairs. The two exits are both locked from the same side.
Once you see it, the arithmetic stops being mysterious. Commitments accumulate at a level that can create them and cannot dissolve them. The permanent capacity shortfall that every PM I know treats as a planning failure, or as evidence of their own inadequate prioritization, is not a failure at all. It is the predictable output of a structure where the right to add load is distributed and the right to shed it is not.
What the comp process proves
There is a more honest version of “I’ll get back to you,” and occasionally someone says it: that’s above my pay grade. I used to hear that as an excuse. It isn’t. It is an accurate statement of decision rights. The difference between the two sentences is not sincerity. It is information. One tells you where the decision isn’t. The other tells you nothing at all, which means you can’t route around it, can’t follow up on it, and can’t tell the difference between a request being considered and one that evaporated in a corridor.
Compare this to asking for a raise, which looks like the same structure and is not. Your manager can’t approve it either. They also say they’ll take it further. But nobody experiences that as a black hole, because the comp process is instrumented: there is a cycle, there are bands, there are criteria, and there is a month in which you find out. You may hate the answer. You will get one.
That is the whole difference, and it disproves the easy conclusion. Centralization is not the defect. Comp is centralized and it works well enough that we barely discuss it. What breaks is centralization without a published path.
The mirror of accountability laundering
I wrote a while back about how accountability gets laundered, how ownership of a decision dissolves through a chain of individually rational steps until no one holds it. This is the mirror image, and in some ways the worse one. Here, nothing dissolves. Authority moves up, cleanly and deliberately, and accountability stays exactly where it was. The team lead who can’t change the load still owns the delivery date. Still owns the incident review. Still writes the performance assessments for people whose workload they didn’t set and can’t reduce. Nobody laundered anything. The two things were simply unbolted from each other and left in different rooms.
I have watched what that does to someone over a few years. When every question about load has been answered upstairs for long enough, upstairs becomes the answer to questions that were never about load. Should we sequence this before that: let me check. Do we accept this technical approach: I’ll raise it. Rights that were never taken away stop being exercised, because exercising them stopped being the reflex. If you know a manager who appears to decide nothing at all, I would look at what they were allowed to decide five years ago before concluding anything about them.
What to do differently
The first part is small enough to do this week.
Take your last ten asks to whoever you report to. Not the big ones: all of them. Mark each one: decided here, or relayed. You will find a clean split, and it won’t run along the line you assumed. It runs along load. Everything that changed the size of the commitment left the building; everything that changed its shape did not. That map is worth having, because most of the frustration I used to carry into those meetings came from sending load-shaped asks to a node that only holds shape rights. It was never going to work, no matter how good the business case was.
Then ask for the instrumentation. Not for the headcount: for the process. Who decides this. Against what criteria. In which forum. By when do I hear. Ask it in writing, politely, once.
You will probably not get an answer. That is not a failed attempt; that is the finding. If nobody in the chain can name the owner, the criteria, or the date, then what you’ve been calling an escalation path is a queue with no service level and no guaranteed exit. Knowing that changes what you do next, because you stop treating “I’ll get back to you” as movement and start treating it as a null result. It also gives you something specific to say when the people who do hold the rights are present: not we need more people, which is a request, but there is no defined path for load decisions in this domain, which is an observation about their process.
And while you wait, decide the thing you can decide. Not by asking permission to cut scope, but by publishing what isn’t being delivered and letting the objection come to you. I have argued before that a priority list is a protection mechanism, that it gives a team explicit permission to let things die. That argument assumed the right to let things die was yours to exercise. Where it isn’t, the list does different work: it stops being a permission artifact and becomes an evidence artifact. Here is what we committed to, here is what capacity allows, here is the difference, here is who has been informed. You aren’t deciding to decommit. You are documenting that no one has, and putting the consequence where it can be seen.
There is a third option: route the ask past the relay entirely and put it in front of the people who hold both rights. In some organizations that is unremarkable and in others it ends careers, and no one can tell you from the outside which one you’re in.
But stop optimizing the request. The request was never the problem. It went to someone who couldn’t answer it, and came back as a sentence whose only job was to avoid saying so.