The new game of HR with AI
AI crashes the price of the mechanical part of HR: screening, job-post drafting, review formatting. What starts being worth gold is your judgment about people. And here there's a catch that, in decisions about people, is especially dangerous.
Tuesday afternoon, and a task that always took up the whole morning, reading, cross-checking, summarizing, writing, disappears in a few minutes with AI's help. The feeling is relief. Except, looking closely at the result, there's something odd: a number that "seems reasonable", a cut that rejected an entire group with no explicit reason, a confident claim about something AI never actually saw. The question that separates whoever understood the new game from whoever just used the tool in a hurry is no longer "do you know how to do this fast". It's: "do you know what AI decided behind the scenes, and are you going to let that invisible judgment become a decision without checking it?".
Friday, six in the evening, and you need to close the month's variance analysis and still write the summary for Monday's board meeting. Before, that used to be your whole night: pull the numbers, check them, build the table, write it up. Today AI does the first draft of all of that in minutes, except it also hands you, with the same confidence as always, a cause explanation that "seems reasonable" for a generic deviation, without ever having cross-checked your real supplier who raised prices. The question that separates the expensive professional from the cheap one is no longer "do you know how to build the spreadsheet". It's: "do you know what this number actually means, and what to do with it?".
The client sends the contract at five p.m. on Thursday asking for a review by Friday morning. You ask AI to draft the termination clause, it hands back a polished text in seconds, looking like it knows exactly what it's doing. Except, checking it carefully, the clause cites a legal article that no longer works that way, repealed in a reform AI didn't know about. With the same confidence as always. The question that separates the expensive lawyer from the cheap one is no longer "do you know how to draft the clause". It's: "do you know which legal basis in this clause is real, and which one AI made up disguised as a citation?".
Monday morning, and the quarter's campaign brief needs to be ready for the committee at ten. You ask AI for the copy and the media plan, it delivers everything formatted, with a detailed persona of the target audience. Except that persona carries an outdated stereotype, the "urban young adult" that ignores the entire real audience from smaller towns that actually buys your product the most, because AI's training data comes from a different context. The question that separates the expensive marketer from the cheap one is no longer "do you know how to build the brief". It's: "do you know if this persona is actually yours, or is it the generic stereotype AI repeated?".
Friday, six in the evening. You need to close the screening for a competitive opening, some two hundred resumes stacked up, and still deliver feedback on three performance reviews by Monday morning. Before, that ate the whole weekend: reading resume after resume, cross-checking criteria, writing each feedback paragraph slowly so it wouldn't sound cold. Today AI does the first cut of the screening and drafts the three pieces of feedback in minutes. You breathe a sigh of relief, save the file, almost close the laptop. Except, looking again at the list of "best fits" it handed you, something bothers you. The rejected resumes have a pattern. Foreign-sounding names, universities outside the traditional circuit, a neighborhood that isn't one of the "good" ones on the resume. No explicit rule says this. AI just learned it, and decided, without telling you. The question that separates the expensive HR professional from the cheap one is no longer "can you filter two hundred resumes in a day?". It's: "do you know why this criterion rejected exactly this group, and are you going to let it decide who joins the company without auditing it first?".
Monday morning, roadmap prioritization meeting, and before that you spent hours tabulating user feedback, building the PRD, formatting the slide with the quarter's metrics. Today AI groups the tickets, drafts the PRD, and builds the chart in minutes, but it also assumes, with the same air of certainty, a retention rate that "seems reasonable" for a similar feature, without having looked at your real analytics dashboard. The question that separates the expensive PM from the cheap one is no longer "do you know how to document". It's: "do you know which problem is worth solving now, and which number in this PRD came from your data and which came from AI's guess?".
Friday afternoon, and you need to update the forecast and send three pending proposals. Before, that used to be your whole night: reviewing the pipeline, building the proposal, writing the follow-up email. Today AI generates the draft of all of that in minutes, but it also applies, with the same confidence as always, a proposal-to-close conversion rate that's a market average, not your real funnel's. The question that separates the expensive salesperson from the cheap one is no longer "do you know how to build the proposal". It's: "do you know which account is actually going to close, and which number in this forecast is yours and which one AI made up on top?".
Start of shift, and you need to understand why the SLA blew up yesterday and adjust the operation for today. Before, that was hours cross-referencing delivery spreadsheets, building the dashboard, writing the incident report. Today AI pulls the data, builds the dashboard, and drafts the report in minutes, except it also estimates, with the same air of certainty, a damage rate that "seems reasonable" for a healthy operation, not for yours, which is running worse. The question that separates the expensive operations professional from the cheap one is no longer "can you consolidate". It's: "do you know where the real bottleneck is, and which number in this dashboard actually matches your process?".
The day before an audit, and you need to close the risk map and review whether the data-privacy policies match actual practice. Before, that was days reading control after control, building the matrix, formatting the opinion. Today AI scans the documents and drafts the matrix in minutes, but it also assumes, with the same confidence as always, an enforcement probability that "seems moderate" for a generic case, without knowing about the notices you already received this year. The question that separates the expensive compliance professional from the cheap one is no longer "do you know how to survey the controls". It's: "do you know which risk is actually material, and which number in this matrix goes into the committee minutes without you having checked the source?".
Friday night, and there's a stuck deploy and a production incident to postmortem. Before, that was the whole small hours reading logs, building the diagnosis, drafting the fix's code. Today AI pulls the logs, suggests the patch, and drafts the postmortem in minutes, but it also points to, with the same air of certainty, a root cause that "seems plausible" for any similar stack, without having cross-checked your real deploy history. The question that separates the expensive dev from the cheap one is no longer "do you know how to write the code". It's: "do you know what the real root cause is, and which line in this diagnosis you need to check before you sign off on the postmortem?".
End of sprint, and you need to synthesize the research interviews and adjust the prototype's flow before usability testing. Before, that was hours transcribing, grouping findings, redesigning the screen. Today AI summarizes the interviews and drafts the clusters in minutes, except it also assumes, with the same confidence as always, a drop-off point that "seems obvious" for any generic signup flow, without having seen your latest real user research. The question that separates the expensive designer from the cheap one is no longer "do you know how to organize the research". It's: "do you know where the user actually gets stuck, and which finding in this summary came from your interview and which one AI generalized?".
Sunday night, and the board wants the expansion analysis ready for Tuesday. Before, that used to be the whole weekend cross-referencing competitor data, running scenarios, building the deck. Today AI drafts a first version in minutes, TAM, SAM, SOM, and adoption curve all ready. The problem is it also hands you, with the same air of certainty, a penetration rate it made up because it "seems reasonable" for a generic market, without ever having seen that your last launch converted a lot less. The question that separates the expensive strategist from the cheap one is no longer "do you know how to build the deck". It's: "do you know which number is actually yours, and which one AI guessed dressed up as analysis?".
Let me start with a somewhat uncomfortable truth about HR. A good chunk of what used to take up your whole day, reading resume after resume, cross-checking criteria, writing each piece of feedback slowly so it wouldn't sound cold, AI now does in minutes. Think about it: does that scare you or free you? The right answer depends on you understanding which part of your work became a commodity and which part became worth more. And there's a second question, more serious than that: when what's at stake is deciding about a person, how much can you trust what the machine hands you without looking under the hood? This lesson settles that account.
The core idea of this lesson. AI crashes the price of the mechanical part of HR: resume screening, job-post drafting, performance-review formatting. That's good news, because more of you is left over for the part AI doesn't do: understanding someone's real fit with the team, deciding to hire, promote, or let go, having the hard conversation, communicating it with humanity. But HR has a catch other areas feel less: here the decision is about PEOPLE, and that amplifies AI's two worst habits, bias that discriminates by proxy and hallucination that invents information about a real person. That's why auditing what it delivers before deciding about someone isn't a luxury, it's part of the job. This track teaches you to operate in this new game.
01What AI commoditizes (and why that's good news)
Let's call it what it is. These things, which used to take up your hours, AI already does fast and cheap:
- The first cut of resume screening, separating who clearly doesn't meet the criteria from who deserves a closer look.
- The draft of the job posting, the feedback email, the climate-survey summary.
- The formatting of the performance review, the headcount report, the onboarding plan.
- The first read of a recorded interview or an exit form.
The big issue here is simple: when mechanical work gets cheap, it stops being your edge. The professional who only knew how to filter resumes fast loses value. That sounds harsh, but there's another side, and it's the good side.
What shifts in your day:
Here's the good news: the time you used to spend reading resumes one by one doesn't vanish, it shifts to the part that's worth more, truly understanding who has fit with the team. Whoever understands this stops fearing AI and starts using it to climb altitude. Fair enough?
02What's still yours (and got more expensive)
Here's the part AI doesn't do for you, and precisely because of that, it's now worth more:
- The real fit. AI points out who matches the job posting's keywords, but whether that person will thrive with that manager, that team, that moment in the company, that's reading people, and it's yours.
- The "so what?". The dashboard says turnover rose three points on the sales team. So what? What does that mean, what do you do about it. AI describes; you interpret and decide.
- The decision. AI recommends who to hire, promote, or let go, but whoever signs off, and answers for that person afterward, is a human being. Always.
- The hard conversation. Giving tough feedback, communicating a termination, mediating a conflict between teams. That's humanity, and humanity doesn't commoditize.
Notice that all of this is judgment about people, not screening. And judgment about people is exactly what gets expensive when screening gets cheap. From here on, your value lives much more in "does this person fit this team, and what do we do about it" than in "I managed to filter two hundred resumes".
03HR's deadly catch: bias and hallucination in decisions about people
Now the part I can't let you forget, because in HR it costs a lot more than money. AI learns historical patterns, and a historical hiring pattern carries the biases of whoever hired before. When you ask it to rank resumes, it can learn, without anyone programming it on purpose, that a certain name, a certain university, a certain neighborhood "fits less" with the opening. That's not opinion, it's discrimination by proxy, hidden behind a number that looks neutral.
This has actually happened. In 2018, Reuters revealed that Amazon scrapped an internal AI screening tool because it had learned to penalize resumes containing the word "women's", as in "captain of the women's chess club", simply because the company's hiring history was mostly male. And it isn't a thing of the past: a 2024 study from the University of Washington (Wilson & Caliskan) showed language models preferring names associated with white men in about 85% of resume-ranking tests. Bias doesn't announce that it's there. It comes with the same confidence as a correct result.
The second risk is hallucination: AI invents a "red flag" in someone's history, cites a reference it never checked, states something about a real person that simply isn't true. In marketing, a shallow piece of text slides by unnoticed. In HR, a made-up piece of information about a person can cost them the job opening, the promotion, the employment.
That's why the golden rule of this track, which you'll see repeated in every choreography: in decisions about people, every AI output goes through a bias audit before it becomes a decision. It's not distrust of the tool, it's professional hygiene. AI speeds up the draft; you make sure the decision about a person didn't come from a hidden pattern.
04The map of this track
This lesson was the framing. The rest of the module is hands-on, installing systems into your real work, one at a time:
- Connecting AI to context about people, with LGPD doubled up because here the data is personal.
- The choreography of bias-free screening, with explicit criteria beforehand and a calibration audit afterward.
- The job posting that attracts the right candidate, without language that turns away who should apply.
- Onboarding that scales, personalized by role.
- Listening to the organization's climate and signals, without turning into surveillance.
- The audit of a decision about people, the checklist before any termination or promotion.
- And, at the end, your HR OS: the library of prompts, models, and agents that keeps working for you.
Each one takes a task you already do and redesigns it. By the end, you won't have learned about AI, you'll have installed AI into your way of working, with the care a decision about people demands. Shall we?
Do it now
Take an HR deliverable you did this week, your real task or another one (a screening, a review, an onboarding plan). Grab a sheet of paper and split it into two columns:
- Mechanical (what AI would do for you): the first cut, the formatting, the draft of the text.
- Judgment (what's still yours): the real fit with the team, what the number meant, the decision, how you communicated it.
Now look at the proportion. How much of your time went to the left column? That's exactly the time this track is going to give you back, for you to spend on the right column, the one that actually decides about people.
Practice
1. In the new game of HR, what loses the most value as a professional differentiator?
2. Why is a bias audit especially non-negotiable when AI decides about people?
3. What does the case of Amazon's screening tool, revealed by Reuters in 2018, illustrate about HR's deadly catch?
</content>
For the board
On what got cheapreading CV after CV, cross-checking criteria, drafting feedback. Necessary, and no longer a differentiator.
On the cost of errorin decisions about people, the error does not only cost money. It costs someone's livelihood.
On biasit does not announce itself. A skewed history becomes a criterion wearing the face of objectivity.
Thanks for the feedback. It helps sharpen the next lesson.