Hiring your first employees as a startup founder

Hiring your tenth employee is a process problem. Hiring your first is a judgement problem, and everything that usually helps is missing.

You have no baseline: nobody has done this job here, so you cannot compare a candidate against a colleague who is good at it. You have no interview loop, so whatever you ask is whatever you thought of that morning. You are the only interviewer, which means your first impression goes unchecked. And the stakes are lopsided. The first hire sets the standard the next five are measured against, does work nobody can cover, and is painful to unwind in a company of three.

The answer is not a longer process. It is a narrower one: decide in advance what the role has to produce, test that specific thing the same way with everybody, and write down what you saw before you discuss it with anyone. What follows is that, in order.

Most first job descriptions are a list of everything the founder would like to stop doing. That reads as a request for a person who does not exist, and it leaves you nothing to interview against.

Write instead what you want to be true ninety days after they start. Three or four lines, each one something you could check.

• "The onboarding email sequence exists and is sending, and I am not writing it."
• "Support replies go out within a day without me in the inbox."
• "The billing bugs in the backlog are closed, and new ones get fixed without me."

Then say what they will inherit: no documentation, a codebase one person understands, a product that changes weekly. That belongs in the advert, because the candidate who is put off by it would have left in month four anyway.

Finally, write down the one thing this role is not. First hires drift into everything, and "this is not a sales role" in the advert saves you both a month.

Pick three things, four at most. Beyond that you will test none of them properly and decide on impressions instead.

• Can they do the core task? Not their whole resume: the one thing the ninety day list depends on. Where the job is making something, a work sample tests this better than any conversation.
• Do they act without a manager? There is no process to follow at your company and nobody to ask. Ask about a time they decided alone, and what happened when the decision was wrong.
• Can they explain their work to somebody who was not there? They will do this with you every week, and with the people who buy from you.

Write one or two questions for each, and ask everybody the same ones. The comparison at the end is only as good as the overlap between the conversations.

Founders usually interview too few, hire the best of three, and never learn what the fourth looked like. Two more conversations cost two hours; the wrong first hire costs months.

1. Everyone who applies gets a structured first round. At ten applicants you could do these yourself. Past twenty or thirty you will not, honestly, and the ones you skip will be skipped by the order they arrived in.
2. Six to ten get a proper conversation with you, against the same three things, in the same order.
3. Two or three do a small paid work sample, scoped to a few hours, that resembles the actual job.
4. Two references for whoever you are about to offer, asked specific questions rather than what they were like.

If fewer than six people reach step two, the problem is the advert or its reach, not your process.

The hard part of a first hire is that you are the instrument. Three habits make that measurably better.

Score before you discuss. Write your judgement on each of the three things, with the evidence, before you talk to a co-founder or read anyone else's notes. Otherwise whoever speaks first sets the answer.

Keep the structure identical. The re-analysis of personnel selection research by Sackett, Zhang, Berry and Lievens revised the whole field's validity estimates downwards and still found structured interviews top-ranked among the procedures it examined. Structure is not a formality; it is most of the reason an interview predicts anything at all.

Distrust your own override. You will want to forgive a weak answer because you liked somebody. In Hoffman, Kahn and Li's study of job testing across fifteen firms, managers who hired against the test's recommendation ended up with worse hires on average. Override it if you must, but write down why, so you can check yourself in six months.

The founder-specific version of this trap is hiring the person most like you. It is comfortable, it interviews beautifully, and in a company of two it doubles your blind spots instead of covering them.

The help is narrow and real: you stop being the calendar. Every applicant gets the same first conversation without you booking it, sitting in it or writing it up, which is what makes it affordable to let everybody have one instead of picking six resumes and hoping.

On Career1, posting the role gives you a job page, and applying starts the interview. The AI interviewer speaks with each applicant in the browser for about eight minutes, builds its questions from your role and their resume, and follows up when an answer is thin. Back comes a score against the role, scored sections with written reasoning for technical depth, communication, problem solving, experience fit and culture fit, a ledger of which claimed skills were actually demonstrated, risks, and questions worth asking in your own round. The transcript and recording sit with it, and there is a full example to read before you trust one.

What it does not do is the judgement. It does not source candidates, does not run your work sample, does not take references, and rejects nobody. For a first hire the conversation with your final three is the entire decision, and it stays yours.

• Hiring a title instead of a job. A "Head of Growth" as employee number one usually means the work was never written down.
• Testing enthusiasm. Everybody interviewing at a three-person company is enthusiastic. It separates nobody.
• Skipping the work sample because you liked them. The work sample earns its keep precisely when you already want to say yes.
• Unpaid work samples, or long ones. A few hours, paid, scoped. More than that loses you every candidate who has options.
• Hiring for the company you expect in two years. Hire for the ninety days you wrote down.
• Never sending the ninety day list. If it was good enough to hire against, it is good enough to give them on day one.

• You have no applicants. Career1 interviews the people who apply to you; it does not find anybody. Reach comes first, and nothing here substitutes for it.
• The job is making something. Use a work sample. An interview tells you how somebody talks about work, not what they produce.
• You are hiring a co-founder or an executive. That is a relationship decided over weeks of conversation, not a screen.
• You interview in a language other than English. The interview is English only at the time of writing.
• There is one candidate and you already know them. Talk to them.
• You need an applicant tracking system. Career1 is not one; the difference is in ATS or AI interviewer.

1. Create a workspace: an email address and a password, no card.
2. Post the role, using your ninety day outcomes as the description.
3. Put your two or three must-have questions on the application form.
4. Share the job link where the people you want actually read.
5. Applicants interview themselves, at any hour. You read reports and choose who to meet.

The free plan vets two candidates a month and needs no card, which is enough to judge the reports before committing to anything. Prices for more volume are on the pricing page. If the applications arrive in the hundreds, the process for that is in screening 200+ applicants.

Both sources below were opened and checked on 26 September 2026.

• Sackett, Zhang, Berry and Lievens, "Revisiting meta-analytic estimates of validity in personnel selection", Journal of Applied Psychology: doi.org/10.1037/apl0000994.
• Hoffman, Kahn and Li, "Discretion in Hiring", NBER working paper 21709: nber.org/papers/w21709.