Before You Answer Anybody Else

Nobody in your building has asked what you think about this. They have asked you what the rules are.

That is how most things reach HR: as somebody else’s question, in a doorway. A hiring manager forwards a vendor page and wants to know whether the screening add-on is allowed. An employee closes your door and asks whether they should be worried, and is not joking. Then a leadership email arrives in the tone people use for small favors: we should probably have a policy on this, could you take a look.

Underneath all of them sit the same two stories everybody else is hearing.

The first one ends the function. Postings write themselves, applicants get ranked overnight, and what is left of you is a dashboard and a lawyer. It usually turns up as a chart with job titles down one side and a shaded band at the bottom. Yours is in the band.

The second story comes with a login. An AI panel appeared inside your ATS over a weekend, switched on by default, promising to halve your time-to-fill. There is a booth at the conference and somebody there will explain it to you slowly. Thirty days free, card details now.

The two stories are built the same way. Both are the size of an industry, both sit past next year, and neither hands you a thing to do on Thursday. Pick a side over lunch if you want one; nothing on your desk moves either way. The req that opened Monday still has no posting. Open enrollment is six weeks out and the carrier summary is thirty pages of a language nobody speaks. And somewhere in your system are forty people whose status has not changed since April, every one of them still checking.

Neither story is the business of this book. The forty people are.

The first story deserves a straight answer, and the answer is a list. Every item on it is something you did this month with a person in front of you.

It cannot sit in the room. When a role goes, somebody has to sit three feet away and say a sentence that gets repeated at a kitchen table that night. Then stay in the chair. Answer a question about health coverage the person is not really hearing. Walk them out at their pace, not the building’s, and still be there twenty minutes after the meeting should have ended, because the meeting was never the point.

It cannot tell you whether somebody is lying. The answer is good: structured, specific, a number in it. Then you ask what they would do differently, it lands half a beat late and a shade too smooth, and you go and check the second reference properly. That is not information. That is what four hundred interviews left behind in you.

It cannot tell you whether a complaint is a pattern or two people who cannot stand each other. Both can describe the same meeting truthfully and describe two different meetings. What decides it is the thing you carry and never wrote down: this is the third person off that team in eighteen months, and two of them used the same word on the way out.

It cannot hear what a room declines to say. A manager tells you somebody is not a culture fit and looks slightly past you while he says it. You now know something you would rather not, no page of the handbook covers it, and it has to be dealt with this week.

And it carries none of it. Your name is on the offer letter. Your name goes on the position statement if a charge ever lands. Nobody deposes software.

In your work that last part is not only practical, it is written down. Using software to screen, score, rank or select people is regulated activity now, with bias-audit obligations and notice rules in a growing list of places. Chapter 10 sets it out plainly. The line between drafting a document and deciding about a person is a legal line, and this book stays on the safe side of it.

It does not know your employment law either. Ask when a final paycheck is due where you are, or how many hours make somebody leave-eligible at a company your size. Out comes a tidy, confident paragraph describing somewhere else, or nowhere. Anything it says about employment law starts a question for counsel. It never ends one.

So: no. Not the room, not the read, not the decision, not the exposure.

There is a real thing under that worry, though, and it is not the thing in the chart. The risk in your building is a screening feature switched on by somebody in another department who has never heard of a bias audit. Which is an argument for reading on: the person who ought to be writing your company’s rules is the one who understands both the tools and the risk, and Chapter 9 makes that project yours.

Now the uncomfortable half, since pretending nothing was moving would fall apart by Chapter 4 and take the rest of the book with it.

Something is moving, and it is not the room. It is everything written around the room.

Nobody warns you, when you become the whole department, that the job is mostly composition. You write the req, then the posting. The offer letter. The rejection, or forty of them, or none. The policy announcement, then the FAQ, then the manager brief so six managers stop giving six different answers by Wednesday. The benefits explainer that turns thirty pages of carrier English into something a person can choose from. The handbook section nobody has opened since 2019. The note written the same day it happened, dated and factual, because in a year it may be the only thing between the company and somebody else’s memory. And the page of talking points for Friday that you read again in the car.

None of it is judgment. It is the paperwork judgment arrives wrapped in. Nearly all of it happens after ten to five, or never.

That pile is the one thing these tools do well. Not deciding whether to open the req; the posting. Not deciding that somebody’s role is going; the clear, humane page that keeps Friday’s meeting short, so your attention is on the person and not on your own sentences.

They fail in ways worth naming, because here the failures get expensive. The main one is that they invent, and nothing in the tone changes when they do. A leave-eligibility threshold. A COBRA window. A benefit amount your carrier has never offered. A degree requirement in a posting where you mentioned no degree, because that is what postings looked like where it learned to write them. And it will offer to go further than you asked, cheerfully, mid-task: shall I rank these for you? No. That is the regulated line, and holding it is your job.

Every workflow in here is built on the same split. The facts, the numbers, the law, the names and every decision about a human being stay with you. The sentences are its half. Where the split is easy to drop, a WATCH OUT box names the specific failure and what to look at. Looking takes about a minute.

Take an ordinary Thursday, the kind that does not look like much from the hallway.

By twenty to ten you have answered four questions in a doorway. Whether bereavement leave covers a grandparent. Whether Monday’s start date can move. Whether somebody can take time they have not accrued. And one that opens with “this is probably nothing, but,” which is the only one that matters and the only one with nowhere to sit.

At ten a hiring manager needs a req open by Friday and the posting up today. So you take the last posting for something similar, change the title, and inherit six requirements nobody alive can justify, including a degree riding along since before you started.

At eleven you build an offer letter slowly, because that document is a promise with a signature under it. At one, a manager tells you he wants somebody gone by Friday. Nothing is documented and nothing has been said to the person. That conversation takes fifty minutes, and nobody will ever see it, because what it produced is a thing that did not happen.

At ten to five you write the posting. Eleven minutes, because eleven minutes is what is left. It is fine. Fine gets you a pipeline that looks like the last pipeline.

At ten past six you open the folder with forty people in it, meaning to do this properly. Three come out well. Then the sentences stop, because writing a kind no takes something particular and you ran out of it around four. The other thirty-seven get nothing again, and one of them is telling somebody tonight what it is like to apply at your company.

None of that is a skill problem. Three of those rejections were good, and the fifty minutes at one o’clock were the work of somebody genuinely good at this. It is a counting problem — doing any of it properly costs twenty minutes, the day asked for thirty of them, and the department is one person.

So here is the book, stated plainly.

It aims a plain AI chat tool at the writing that surrounds HR and recruiting. Any of the big ones will do. Postings built from a real intake, outreach that does not read like a blast, structured interview guides, and a rejection kit that clears the debt. Policy announcements people read, benefits explainers that produce decisions instead of tickets, onboarding sequences, the all-hands FAQ. Handbook sections drafted for counsel rather than around counsel, SOPs, the review-season structure pass, offer letters from your approved template. Termination talking points, the PIP, accommodation scaffolding, documentation that stays factual. Law updates distilled, comp organized. And your own career, including the policy your company has not written yet. The prompts run to 150, and they sit again in a numbered library at the back that you can search in ten seconds while somebody waits in your doorway.

Each of the 150 was run in all three tools, and the ones that worked in only one did not make the book. So this is not a forecast of what AI might eventually do for HR. It is the wording that produced something sendable.

Things it is not: legal advice, a stand-in for employment counsel, or anything that screens, scores or ranks a person. Nothing in here puts a machine in a position to select a human being. Plenty of this work sits near expensive things: employee and candidate data, anything medical, comp with a name attached, an investigation, a document that could become an exhibit. Wherever it does, a RULES CORNER box shows you where the fence is. Chapter 10 collects them. Read it before any of this reaches a real candidate.

No technical skill is required. A phone, an email address, half an hour. Chapter 2 handles the setup. Chapter 3 is about the one thing you have to be able to do, and you already sell it: briefing one of these tools is a hiring-manager intake. What the job does day to day. What is a must-have and what is a wish. Run an intake that comes back with a job instead of a wish list and you can do this. Not a compliment, a description of how the thing works.

A last word, about scale.

It does not change your life, and on its own it does not change your title. It changes the hour after five, and within a month, most evenings of the week. The table overleaf splits those hours by chapter and totals about eleven in an ordinary week. That figure is real and it is not a gift: week one takes time and returns none, as learning anything does. The balance tips around week three.

Two habits is what most readers walk away with, and two is what this was built for. Usually the rejection kit and the posting workflow. Run them weekly until they stop announcing themselves as something new, leave the career chapter shut until spring, take four hours back, and stop thinking about it. The test is not whether you reached Chapter 12. It is whether two things stuck.

Nobody is coming for the room. Somebody with judgment still has to sit three feet away on a Friday morning, hear what a manager is not saying, and work out whether a complaint is a pattern or a personality. What is genuinely available here is smaller than either story and far more useful: the posting written from an intake instead of at ten to five, forty rejections owed becoming none, and a Thursday that ends when you close the laptop.

Read the table. Then give Chapter 2 half an hour, and go and clear the forty.