Every article on this subject opens with a salary range. That number is the least useful part of the decision, because it prices a person when what you are buying is the closing of a leak.
This is the buyer's version: what actually moves the cost, what is usually left out of the quote, and the comparison worth making.
What moves the number
Scope. Executing a process somebody else defined is one price. Owning the outcome of that process is another. The second includes deciding what to stop, and that is where most of the value sits.
Seniority, measured in judgment rather than years. The expensive skill is not building the automation. It is knowing what not to automate: what should stay manual until the process is written down, what needs a human check because it is irreversible, and which data cannot be trusted yet. Cheap hires build everything you ask for.
Time zone overlap. This is the real reason Latin America is on your list. Most of the region overlaps the US working day, which means questions get answered the same day instead of the next one. That overlap is worth more than a percentage point of rate.
Capacity versus ownership. Hands for a known process, or a person accountable for a number. These are different purchases and they should not be compared on the same line.
Engagement shape. A defined project, a monthly retainer, or a full time hire. Each carries a different risk profile for you: a project caps your exposure and your depth, a hire maximises depth and your commitment.
Scarcity of the specific skill. Analysis of job postings by Lightcast found that roles requiring AI skills advertise materially higher salaries than comparable roles without them. The World Economic Forum's outlook points the same way on demand for technology and AI skills. Treat both as a market signal, not a quote.
What is usually missing from the quote
Four costs that appear after the contract is signed and rarely before:
- Ramp. The time before anything ships. Ask what it is and what shortens it.
- Access and tooling. Seats, API keys, a sandbox. Small money, large delays when unplanned.
- Your team's time. Every question they ask is time from someone who was doing something else. A good hire asks a lot of questions early and few later.
- The handover. What happens if this ends. If there is no runbook, the real cost of the engagement includes rebuilding it.
The comparison that actually decides
Not one salary against another. The person against the problem.
Price your leak first. If inbound leads sit for six hours before anyone calls, work out what that costs: the conversion difference between a fast first contact and a slow one, times your volume, times your average deal. If your reps spend the first part of every call looking things up, price that hour. If nobody can trace which campaign produced revenue, price the budget you cannot defend.
Most companies that do this arithmetic find the hiring question answers itself, in one direction or the other. And the ones who find the leak is small have learned something more valuable than a salary range: they did not have the problem they were about to hire for.
What to ask for instead of a rate card
Ask for the work. A profile that shows what someone shipped, in which context, with what result, tells you more about price than any benchmark, because it lets you judge whether they solved your problem or an easier one.
Then structure the risk: a small paid project against a real problem before a long contract. It costs a fraction of a bad hire and it answers the only question a rate card cannot.
See the work before you price it
MitHub keeps open profiles with the deployments written out, split by whether the person faces your client or builds inside your systems. Browsing and contacting them is free, and you pay a fee only when you hire, which we reinvest in the talent's education rather than keeping.