When to Hire Your First Employee in an AI Services Business

Hire your first employee for an AI services business when you have turned down work you could have delivered profitably, for at least three consecutive months, and you hold enough cash to pay that person for six months without a single new contract signing. Both conditions. One without the other is how solo operators end up with a payroll obligation and a sales problem at the same time.

Everything below is the arithmetic behind those two conditions, and the alternatives that are usually better first. If the underlying question is really how the business makes money at all, the monetisation options open to AI builders is the better starting point.

The signal is refused work, not busy-ness

Being busy is not evidence that you need a person. Being busy is evidence that you are busy, which can equally mean your pricing is too low or your delivery is inefficient.

The signal that actually justifies a hire is demand you declined. Keep a list. Every time you say no to an enquiry, write down the date, what it was, what you would have charged, and why you said no. After three months you have a real number rather than a feeling.

Then read the "why" column, because it tells you what to do next:

|---|---|---|

Only the first row points at a hire. That distinction matters because a hire is the most expensive and least reversible response to overflow, and the other four rows have cheaper answers.

What it costs to hire a first employee in an AI services business

The salary is not the cost. In most of Europe and the UK, loaded cost lands somewhere between 1.2 and 1.4 times gross salary once employer contributions, pension, insurance, equipment, software seats, and holiday cover are counted. The UK government's guide to employing staff for the first time lists the statutory obligations that make up most of that gap, and it is worth reading before you commit to a number. Add unbillable time: a new hire in a services business is rarely above 60% billable in the first quarter, and expecting more is how people end up disappointed in a perfectly good employee.

A worked example, using round numbers you should replace with your own:

• Gross salary: 55,000 a year

• Loaded cost at 1.3x: 71,500

• Realistic billable hours in year one: 900 (allowing ramp-up, admin, and holiday)

• Break-even rate: about 80 an hour before any profit

Now compare that break-even to what you actually charge and collect. If your effective realised rate is 95 an hour, this person contributes roughly 13,500 in year one, before any of your time spent managing them. That is a thin margin for a large commitment, which is not an argument against hiring. It is an argument for hiring when your rate has room in it, which is why pricing an engagement properly comes before headcount, not after.

The six-month cash rule

Hold six months of that loaded cost in accessible cash before the start date. Not projected revenue. Not a signed pipeline. Cash.

The reason is specific to services businesses: your revenue and your ability to sell are the same resource, which is you. The month you start managing someone is the month your own selling time drops. If the pipeline thins at the same moment payroll starts, you need runway that does not depend on you fixing it immediately.

If you cannot model whether you can sustain this, a simple forward projection is worth building before the hire rather than after. Forecasting revenue for a subscription product uses the same mechanics even when your revenue is project-based: known contracted work, probability-weighted pipeline, and a floor scenario where nothing new closes.

Three things to try before hiring anyone

Most overflow does not need an employee.

1. Raise prices. If you are turning away good-fit work at your current rate, that rate is under market for you specifically. A 20% increase both funds and reduces the problem, and it is reversible in a way a hire is not.

2. Use contractors. Variable cost, no notice period, no employer obligations, and a genuine trial of whether you can delegate at all. Many people discover here that their real constraint is that everything lives in their head, which a full-time hire would not have fixed either.

3. Productise. If the overflow is the same engagement repeatedly, the answer may be to stop selling time. Turning repeat work into a productised service raises capacity without raising headcount, and it is the move most AI services businesses reach for too late.

There is also a fourth, which is to narrow. If overflow comes from scattered work across too many domains, niching down reduces the surface area you need to staff at all.

What the first hire should actually be

When you do hire, the instinct is to hire someone like you. In a small AI services business that is usually wrong.

The first hire should absorb the work that consumes you but does not require you. In practice that is delivery execution: implementation against a spec you wrote, evaluation runs, documentation, client communication on live projects. It is not sales, because you cannot yet describe your sales process well enough for someone else to run it, and it is not a more senior version of you, because you cannot yet afford the ramp.

Write the job around the three tasks that eat the most of your week and require the least of your judgement. If you cannot name those three tasks specifically, you are not ready, and the reason is not the money.

What changes the month they start

Two things surprise people, and both are worth planning for rather than discovering.

The first is that your own output drops, and it drops before theirs rises. You are now writing specifications you used to hold in your head, reviewing work you would have done faster yourself, and answering questions at unpredictable moments. Budget for your billable hours falling by a quarter for the first two months, and do not sign a demanding delivery date across that window.

The second is that your documentation debt becomes visible all at once. Everything you know but never wrote down turns into a question, usually at the worst moment. The fix is to treat the first month's questions as a writing backlog: every time you answer something twice, that answer becomes a document. By month three you have an onboarding pack you did not have to sit down and write.

There is also a decision to make early about client-facing exposure. Putting a new hire in front of clients too soon risks the relationship, keeping them hidden too long means you remain the bottleneck you hired to remove. The usual middle path is to have them join calls from week one without owning them, and take ownership of a single low-risk account by month two.

Frequently asked questions

Should my first hire be technical or operational?

Whichever function you are the bottleneck in. Most solo AI consultants are the bottleneck in delivery, not admin, so technical is the common answer. If you are spending eight hours a week on scheduling, invoicing, and client chasing, operational may buy back more capacity per pound.

Is a contractor really a fair test of whether I can hire?

It tests delegation, documentation, and whether your work can be specified for someone else. It does not test retention, culture, or managing someone through a slow month. It is an incomplete test, but it is the cheapest one available.

How long should I wait after the signal appears?

Long enough to confirm the pattern, which is the three months of declined-work data, and no longer. Waiting indefinitely for certainty means you hire during a crunch, interview badly, and onboard someone while you are underwater.

What if my revenue is very lumpy?

Then the six-month rule is a floor and not a target, and a contractor arrangement is probably the right structure for longer than it would be otherwise. Lumpy revenue and fixed payroll is the combination that ends services businesses.

Do I need to hire to look credible to bigger clients?

Sometimes clients ask about team size, usually as a proxy for delivery risk and continuity. That concern is better answered with documented process, named contractor cover, and a continuity clause than with a headcount you cannot sustain.