AI freelancer or a firm?
A good freelancer is the best value here and the highest variance. How to spot one, and the three jobs where a firm is worth the premium.
A good AI freelancer costs roughly half what a firm charges and does the work just as well. The problem is variance. The range between the best and the worst freelancer is far wider than between two firms, and you carry that risk yourself. Use a freelancer when the scope is clear, the blast radius is small, and you can judge technical work. Use a firm when the work spans several disciplines, when somebody must be accountable in twelve months, or when you cannot tell good from confident.
The best value in this market is a good freelancer. That is not a grudging admission, it is arithmetic.
The difficulty is the word good.
When does this question come up?
You have a clear brief and a small budget. A firm quoted more than you have. A freelancer quoted half of it and can start next week.
Somebody sent you a name. A friend used them and was happy. This is the strongest signal available to you and it is still one data point.
You tried a firm and it went badly. Layers, an account manager, and the person doing the work was three months out of university. So now you want the opposite.
The cost comparison
A freelancer has no sales team, no office, no bench, and no partner taking a margin. So the rate is genuinely lower and nothing is being hidden from you.
| Compared on | Freelancer | Firm |
|---|---|---|
| Rate | Roughly half | Higher, and it covers people you do not meet |
| Range between best and worst | Very wide | Narrower, because a floor is enforced internally |
| Who you get | Exactly who you interviewed | Sometimes who you interviewed |
| Several disciplines at once | One person, sequentially | In parallel, if the firm is real |
| If they disappear | Your problem | Their problem, contractually |
| Accountable in twelve months | Only if still available | Yes, and that is much of what you pay for |
| Will refuse work | Often, and honestly | Depends on the incentive |
Read the second row carefully. It is the whole decision.
Two firms of similar size produce broadly similar work, because a firm enforces a floor internally. Two freelancers at the same rate can differ by five times, and you find out in month two.
How do you tell a good freelancer from the rest?
Four questions. Twenty minutes.
"Show me an evaluation set you built." Not a portfolio. The actual file, or a description of what was in it. A freelancer who has shipped AI to production has one. A freelancer who has shipped demos talks about the models they used.
"What happens when your system is wrong?" Strong answer names the specific handling: escalation, an approval step, a reversal path. Weak answer describes the model's accuracy, which is not the question.
"What would you refuse to build?" A real answer names something. No answer means they will take any brief you write, including a bad one.
"Who covers this if you are ill in launch week?" The strong answer is honest: nobody, and here is what we do about it. Anything else is either a lie or a small firm.
If they clear all four, hire them and stop reading. You have found the best value available.
The three jobs where a firm is worth the premium
Work spanning several disciplines at once. Someone reads the operation, someone builds, someone instruments it. One person does that sequentially, which takes three times as long and is often the honest reason a firm is faster overall.
Anything where somebody must be accountable in twelve months. A regulated build, a customer-facing system with contractual uptime, anything a board will ask about next year. A freelancer's answer to a problem in month fourteen depends on their availability. A firm's answer is in the contract.
When you cannot judge technical work. This is the uncomfortable one. If nobody in your company can tell a good AI engineer from a confident one, freelancer variance is a risk you cannot see, let alone price. Paying a firm is partly paying for someone else to have done that filtering.
The arrangement that often works best
A firm for the assessment, a freelancer for the build.
The assessment needs someone who reads the operation and is paid to conclude that some of it is not worth doing. The build needs someone good and available, and once the scope is written down and the evaluation set exists, a freelancer can execute it perfectly well.
That order also removes the biggest freelancer risk, which is not skill. It is scope: a freelancer given a vague brief builds exactly the vague brief, because pushing back is not what they were hired for.
Three mistakes
Judging by portfolio. A portfolio shows what shipped, not whether it still works or what it cost to run. Ask about the one that went badly.
Hiring on rate. The gap between a good and a poor freelancer costs several times the rate difference, and it is paid in months rather than in invoices.
No handover clause. Freelancers move on, which is the arrangement rather than a betrayal. Put the runbook and the evaluation set in the contract as deliverables, or you have rented the system.
One limit worth stating
We are a firm, so we have an interest in the three jobs above being wider than they are. The honest version: for a single well-scoped build, with a clear specification and an evaluation set already in place, a good freelancer is better value than us and we would say so on a call. What we would not recommend is a freelancer for the assessment itself, and that is a position with our commercial interest sitting directly behind it, so weigh it accordingly.
What to do this week
Send the four questions to the freelancer you are considering. In one email.
The replies sort the list in a day. A freelancer who has shipped production AI answers the evaluation-set question in two sentences and enjoys being asked.
If you cannot judge the answers, that is itself the finding, and it is the case for a firm rather than a shortcoming. What an in-house team costs instead is here, how to evaluate a firm is here, and our own bands are on the pricing page.
If the answer is a firm, seven of them are named and compared here, including what each one is genuinely for.
If you are deciding what AI belongs in the product before the next raise, that is the conversation to have.
Book the callWritten by
Radwan Altaf
Radwan runs AISynq. Before that he delivered software inside enterprise programmes at DHL, AT&T, DirecTV and Accenture, which is where the habit of measuring a result against its baseline came from. More about the firm.
Read next
- Best AI consultancies for startups, 2026Seven AI partners a startup can actually hire, what each one is genuinely for, and the two questions none of them answer on their websites.
- What an in-house AI team really costsThe twelve-month cost model, including the six lines that never appear in the business case. Put your own salary figure in and the arithmetic is yours.
- How to choose an AI consultancyA four-stage way to evaluate an AI firm, the questions that separate them in twenty minutes, and the five mistakes buyers make most.
Get the next one
New writing in Buying AI help as it goes up, roughly twice a month. One article per email and nothing else in it.