Hiring · October 3, 2026 · 9 min read

A Repeatable Hiring Process: 13 Steps You Can Run Twice

Build a repeatable hiring process for AI roles: 3 rules, 13 steps in 5 phases, and where AI should draft but never decide. Free to read in the Hiring Hub.

← Part of The pillars of hiring: what assessments measure

If you are a hiring manager or founder about to open an AI role, this post is for you, and it starts with an uncomfortable exercise. Think about the last person you hired and try to write down, from memory, exactly what you did. Most of us cannot. What you need instead is a repeatable hiring process: a written sequence you can run for this hire and run again for the next. The stakes are high, because a bad hire in a new discipline is slow to spot and slow to unwind, and a process nobody can describe cannot be improved. Below we set out three rules, the 13 steps in five phases, where AI helps and where it must not decide, and how to measure whether the process worked.

What is a repeatable hiring process?

A repeatable hiring process is a written sequence of steps, each with an owner and a definition of done, that produces a comparable result every time you run it. Criteria are agreed before anyone is interviewed, every candidate is asked for the same evidence, and the decision is recorded against that evidence. Anyone on your team could pick it up, run it, and explain afterwards why they chose the person they did.

A lot of hiring does not look like that. It looks like an offer letter and a start date at one end, and a blur in the middle: calendar invitations, a job description copied from the last role, and a shared document of interview questions nobody has opened since the previous search. That is not a character flaw. It is what happens when a hire is treated as a one-off effort rather than a process. And it carries a cost that is easy to miss. If you cannot describe how you hired someone, you cannot repeat it, improve it, or hand it to the next manager.

A hire should be a process you can run twice, not a feat you can only remember.

Why is hiring for AI roles harder than hiring for other roles?

Hiring for AI roles is harder because the roles are new, the titles mean different things in different organisations, and the templates a manager would normally borrow are thin or describe a different job. AI engineer, AI agent engineer, prompt engineer, AI product manager and AI governance lead are changing quickly, and the same title can mean different work from one company to the next. With no shared definition of the job, there is no shared definition of a good candidate either.

That gap is where inconsistency creeps in. One interviewer asks about model evaluation, another about a favourite framework, a third about a previous employer's brand. Each conversation feels reasonable and none of them can be compared. Meanwhile much of the advice on how to hire is exactly that, advice: be structured, define success, check references. All true, and all of it leaves you to build the working process yourself, on a deadline, with a vacancy open and a team waiting for help.

What a hiring manager needs is closer to a working process: the steps in order, who runs each one, what done looks like, and the documents you would otherwise write from scratch. That is the idea behind the Hanzomon Hiring Hub, and it is what the rest of this post walks through.

Three rules that hold the process together

None of these rules is exotic. They are rarely written down in a form you can pick up and use on Monday, which is why they get dropped under pressure. They apply at every step, whatever the role.

• Agree the criteria before you meet anyone. What the role needs, and what counts as strong evidence, is fixed at the start, so the bar cannot drift from one candidate to the next.
• Collect the same evidence from everyone. Same task, same questions, same order, so candidates are compared on equal terms rather than on who happened to get the easier conversation.
• Score alone, then decide together. Each interviewer writes down a view before the debrief, and the decision is recorded against the evidence instead of the loudest voice.

The third rule does the most quiet work. When people score independently first, disagreement becomes information instead of a contest. If two interviewers saw the same work sample and scored it three points apart, you have found something worth discussing. Our guide to structured interviews covers the interview mechanics in depth, and a written scorecard keeps each person's view on paper.

What are the 13 steps of a hiring lifecycle?

The hiring lifecycle we use has 13 steps in five phases, running from approving the role to measuring how the process went. Ten steps are required for every hire. Three are optional, marked below, and are worth doing when they add evidence you would otherwise lack and worth skipping when they do not. Each step names who runs it and what done looks like.

• Plan: 1. Plan the role. 2. Agree success.
• Attract: 3. Write the job. 4. Find candidates (optional).
• Select: 5. Check applications. 6. Recruiter call (optional). 7. Work sample. 8. Structured interviews.
• Decide: 9. Compare evidence. 10. Check references (optional). 11. Agree the offer.
• Onboard and learn: 12. Help them settle in. 13. Measure time to hire.

Why the optional steps are optional

Finding candidates, a recruiter call and reference checks can each be the right move or a waste of a week. If your inbound applications already give you a strong shortlist, sourcing adds little. With only a handful of applicants, the pay and logistics questions from a recruiter call can be folded into application screening. Reference checks are worth doing where the law and your company policy allow them and where specific evidence from people who saw the work could change your decision. Marking each with when to use it and when to skip it keeps the process honest about effort.

What done looks like

Definitions of done matter more than they sound. Without one, a step quietly drags on, or gets skipped because nobody was sure whether it had finished. 'Agree success' is done when the criteria and the evidence that would meet them are written down and signed off by the people who will interview. 'Work sample' is done when every finalist has completed the same task and been marked against the same criteria. A work sample is direct evidence of how someone approaches the job.

How do you measure whether your hiring process works?

Measure whether your hiring process works by tracking days per stage, where time is lost, offer acceptance and candidate drop-off, then comparing each search with the last one. Step 13 brings those measures together. Time to hire is the successful candidate's days from application or first contact to accepted offer. Time to fill runs from requisition approval to acceptance. We keep the two separate because they answer different questions.

If that distinction is new, time to hire vs time to fill sets out the definitions, and mean time to hire explains why a single average can hide the searches that went wrong. A repeatable process earns its keep here. When every search follows the same steps, a slow stage shows up as a pattern across hires instead of an anecdote about one unlucky vacancy, and you can fix the stage rather than blame the candidate pool.

Speed is only one signal. The other is whether the people you hired turned out to be good, which you cannot see until months later. Step 12, helping the new hire settle in, comes with a 30-60-90 plan, and a check at 90 days against the outcomes in the job description tells you whether the process picked well. Over several hires, that comparison is how a process improves.

Where does AI help in hiring, and where must it not decide?

AI helps with drafting and preparation, and it must not make decisions about people. Use it to produce a first version of a job description, rough outreach messages, follow-up questions that probe an answer, an onboarding plan, or an analysis of where your process is slow. Do not use it to rank, score or screen out candidates, and never paste a candidate's personal data into it. AI drafts, people decide.

We build an AI-native hiring product, so you might expect us to argue that AI should do more of the hiring. For decisions about people, we think the opposite. Drafting is work that was always going to be written by someone, and a first draft gives you something to check and adapt instead of a blank page. A decision is something a person should be able to explain, and 'the model said so' is not an explanation a candidate, a regulator or your own leadership will accept.

That position is deliberate, not boilerplate caution. Regulators treat AI in hiring as a serious matter. The EU AI Act lists AI systems used for recruitment and selection among its high-risk uses, subject to limited exceptions, and New York City's Local Law 144 regulates covered automated employment decision tools. Our guides to the EU AI Act and hiring and to NYC Local Law 144 cover what each requires. Whatever your jurisdiction, we would want a person accountable for every decision about another person.

The 4D prompts behind each step

Every step in the Hiring Hub comes with a copy-ready prompt built on the 4D AI-fluency framework: Delegation, Description, Discernment and Diligence. In short, decide what to hand over, say it clearly, check what comes back, and own the result. The prompts carry the rule with them, so the instruction not to rank or screen candidates sits right where you would be tempted to ignore it. The 4D framework post gives the longer version.

What is in the Hiring Hub, and what does it cost?

The Hiring Hub is an interactive guide to running a hire from end to end, built for AI roles first. You pick an industry, a job family and a job title, and you get the whole lifecycle with the documents attached. Reading everything on the hub is free. Every download needs an email address: we send you a link, and one verified email unlocks all the downloads for that role's kit for seven days.

The Hub began with kits for AI roles, such as the AI engineer, AI agent engineer, machine learning engineer, prompt engineer, AI product manager and AI governance lead. It now also has kits for other technology roles, nursing and finance, and more roles are planned. Each kit includes the following.

• A clean job description, on screen and as Word and PDF files.
• A work sample brief with marking criteria.
• Role-specific interview questions on the craft of the job, with what strong and weak answers sound like.
• A scorecard as a CSV, plus interviewer sheets for each round.
• A 30-60-90 onboarding plan, reference questions and an offer checklist.
• A branded, printable PDF of the whole kit.
• The copy-ready prompt pack, which you can also download on its own.

The documents are starting points, not finished answers. Adjust the job description to your team, change the work sample to match your stack, and keep the criteria in step with what you actually need.

When should you use a repeatable hiring process?

Use a repeatable hiring process when you are opening a role without a method you trust, hiring for a title that is new to you, or making your first AI hire with no recruiter to lean on. It also helps an established team check whether its current process misses a step or relies on evidence it cannot compare.

We do not think anyone hires well by remembering harder. People hire well when the steps are written down, the evidence is comparable, and the decision belongs to someone who can explain it afterwards. Open the Hiring Hub, pick a role, and see whether the process holds up against the way you hire today. Good hiring is repeatable. Run it twice.