Judgement, at the local rate

I was recently offered PKR 500 an hour to improve a product's visual hierarchy, navigation, interactions and accessibility, and to write down the reasoning behind every change. At the current exchange rate that is a little under two US dollars. A decent coffee in Karachi costs more, and the coffee is not asked to justify its choices in writing.

This is not a post about a rude client. He was not rude. He was thorough, polite and had plainly thought hard about it, which is exactly why it is worth writing about. He priced the work with two discounts that a lot of the market now applies without noticing, one for where I live and one for AI, and both make the same mistake.

Discount one: the postcode

The request was for a "local-market-style" engagement, and for my normal rate for clients in Pakistan. The logic is familiar. Rent is cheaper in Karachi, so the work should be too. It is cost-plus pricing, where the cost is my electricity bill.

The trouble is that the product was never going to ship in Karachi. It ships to his users, in his market, at his prices. A navigation that loses people loses them in his currency. An accessibility gap fails his customers, under his country's rules. The decision I would be making is worth exactly what it is worth to him, and my landlord does not come into it.

The rejected alternative here is not "charge everyone Sydney prices". Geography is a real discount and I would be dishonest to pretend otherwise. A senior designer in Karachi costs less than one in San Francisco, that gap is the whole reason teams hire across borders, and my international clients benefit from it. But there is a difference between less than San Francisco and less than a coffee. The first prices the work and adjusts for the address. The second prices only the address.

A rate is what the work is worth to you, not a reading of my electricity bill.

Discount two: the throughput

The second discount was AI. He would arrive with the requirements, the user flows and a first pass of the screens, all generated, and he assumed roughly three times the throughput for AI-assisted production. To his credit he carved out an exception himself: he was not claiming that expert critique gets three times faster. Then he priced the critique as though it had.

I have actually run the experiment he was hypothesising about. This August I took a finished design for an AI-native business operations platform, more than 190 screens, and turned it into a running application with Claude Code in ten days. Fifteen modules, a documented design system, the lot. Production was fast, faster than three times by any honest measure.

The judgement was not. Most of the real hours went into things no generator offered to do. Modules that had no supplied screens got a deliberately blank route instead of a plausible invented one, because a generated screen sitting beside fourteen real ones reads as real and nobody can find it later. The dark theme turned up a headline that was invisible and a white tile on every empty surface. Nineteen actions turned out to need a confirmation that nothing in the original design drew. None of that was typing. All of it was deciding.

The line I wrote in that project's learnings still holds: building a screen is now the cheap part, so the judgement moved entirely to deciding what not to build. AI made the problems show up sooner. It did not make them cheaper to decide.

AI shortened the part he could see. The part he was paying for stayed the same length.

Two discounts, one mistake

Put them side by side and they are the same move. The postcode discount prices my rent. The throughput discount prices my keystrokes. Both are pricing the input, and neither touches the thing he actually wanted, which he named in his own email: better choices between alternatives, and the reasons for them.

Stack the two and you get under two dollars an hour. Ask what a wrong call costs instead, a checkout people abandon, a form a screen reader cannot get through, a navigation that sends support tickets up every Monday, and you get a very different number. That second number is the one the work is worth. The first one is just arithmetic on my circumstances.

The honest part

I said no, and that has a cost. He was clear there might be occasional work later if the first job went well. Maybe that would have grown into something. I will not find out.

I also know my refusal is not available to everyone. Plenty of very good designers in Pakistan take rates like that because the month needs paying, and the market sets the local number lower partly because enough of us accept it. That is not a failing on their part. It is the reason this deserves to be said out loud by someone who can afford to say no.

And some of the fault is mine. I still quote by the hour, which invites exactly this arithmetic: hours times a local multiplier times an AI discount. If I priced the decision instead of the time, half of this conversation would never have started. I am working on it. Old habits bill hourly.

What I would tell anyone hiring across a border

Pay for the decision in the currency of the market it ships to. Before you apply a local rate, ask what a wrong call would cost your product, in your market, with your users. Before you apply an AI discount, ask whether the part you need, the judgement and the reasons, actually got faster, or just the part you could see.

And to designers working from cheaper cities: quote the work, not your postcode. Your rent is your business. Your reasoning is theirs, and it does not come in a local edition.