Empathy Without Family: What Emma Grede's Leadership Rule Means for AI at Work

Skims cofounder Emma Grede says leaders can be empathetic without treating employees like family. It sounds like a small comment about office culture. It is actually one of the most useful rules anyone has offered for the age of AI at work.

Here is why. Companies are now handing machines tasks that used to belong to managers, writing job posts, screening resumes, building schedules, summarizing feedback, drafting reviews, even coaching bosses through hard conversations. If we cannot explain what real empathy looks like inside a company, we cannot teach a machine to support it. And if we copy the wrong idea of culture into our software, we will scale the wrong behavior to every team at once.

Grede's point cuts through the confusion. You can care about people. You can be flexible, patient, and generous. You can still be the boss. Those two things are not in conflict, and the idea that they are has quietly shaped the way a lot of workplace technology gets built.

The Simple Idea: Care Without Confusion

"We are a family" is one of the most common lines in business. It sounds warm. It often hides something colder.

Families do not fire people. Families do not have pay bands, performance reviews, or quarterly targets. Families do not have an exit interview. When a company calls itself a family, it borrows the emotion of a family without the safety of one. Workers end up feeling like they owe something personal to an employer who can end the relationship with a spreadsheet and a calendar invite.

Grede's framing flips that. Empathy is not the same as closeness. Empathy means understanding what someone is dealing with and responding in a way that respects them. That is different from pretending there is no power gap, no money involved, and no consequences.

A leader can say: "Take the afternoon for your kid's appointment. Also, this project still has to ship Thursday." Both statements are honest. Neither one is cruel. That combination is what real leadership looks like, and it is much easier to define, teach, and measure than "family."

Why This Matters Right Now: AI Has Entered Management

AI tools are already doing management work. Not the big decisions, usually. The everyday ones.

• Software sorts applicants before a human opens a file.
• Systems build shift schedules and predict who can cover a gap.
• Assistants summarize employee surveys and flag "at-risk" teams.
• Chat tools draft performance notes, job descriptions, and policy answers.
• Coaching tools rehearse difficult conversations with managers before they happen.

Every one of these tools encodes a view of what good management means. A scheduling system that ignores rest time encodes a view. A chatbot that answers a complaint with cheerful fake warmth encodes a view. A review generator that rewards people who write in a certain tone encodes a view.

So the culture question is no longer just a human resources question. It is an engineering question. Whatever version of "caring" a company writes into its policies gets copied into its software, and then copied again across thousands of decisions a day.

The Empathy Gap in AI Systems

Machines can be polite. They can be consistent. They can remember every detail of every interaction and never get tired. What they cannot do is care.

That gap matters more than it sounds. People are surprisingly good at detecting fake warmth. When a chatbot responds to a serious complaint with a cheerful "I understand how you feel," trust drops. The worker learns that the system is not listening, it is performing. And once people stop believing the warm words, they stop believing the cold ones too, including the ones that matter, like policy updates and pay information.

There is a second, quieter risk. AI models learn from the text we give them. Much of that text, job ads, handbooks, internal posts, is full of "family" language. Feed enough of it in, and the system starts treating boundaries as disloyalty. Working late looks like commitment. Saying no looks like a problem. Someone who takes all their leave gets quietly scored lower by a tool nobody can see.

That is exactly the trap Grede's advice avoids. If "family" is the goal, then boundaries look like failure. If empathy is the goal, then boundaries are part of the job.

What AI Should Actually Do Instead

Here is the good news. Clear rules are the thing machines handle best. Vague feelings are the thing they handle worst. So the move is to make empathy concrete enough to build with.

That means turning values into behaviors you can see and check:

• Managers respond to time-off requests within two business days.
• Nobody should be messaged outside agreed hours unless it is a genuine emergency.
• Pay ranges are published for every open role.
• Feedback is given in private, praise can be public, and both happen on a schedule.
• Every automated decision that affects a person can be reviewed by a person.

Once you write it down, AI can help enforce it. Tools can flag patterns of overtime on one team. They can remind a busy manager that a review is overdue. They can catch biased phrasing in a job ad before it goes live. They can show a leader that one department always promotes from the same small group of people. They can even check whether pay decisions line up with the published bands.

That is empathy as structure rather than empathy as mood. A worker does not need their boss to feel warm toward them. They need predictable hours, fair pay, and honest answers. Structure delivers that at scale. Good intentions alone do not.

Coaching Managers, Without Turning Them Into Bots

One of the most promising uses of AI at work is helping managers get better at hard conversations. A person can practice a layoff talk, a performance warning, or a conflict between two teammates with a tool that responds the way a real employee might. They can try again. They can hear how they sound.

Used well, this raises the floor. Most bad management is not cruelty. It is avoidance and nerves. Practice helps.

Used badly, it flattens everyone into the same script. Employees can tell when a manager is reading from a template. The words land as distance, not care. The rule of thumb is simple: AI can prepare the conversation, but a human has to have it. Never let a tool deliver bad news on its own. The harder the message, the more it needs a face behind it.

The Hiring Trap Nobody Talks About

Job ads are where culture gets sold before anyone is hired. If a company's posts lean on "family," "always-on," and "we work hard and play hard," the pool of applicants self-selects, and not always in a good way. People who value boundaries read those words and walk away.

Then AI screening tools step in and rank whoever is left. If those tools are tuned toward "culture fit" without a clear definition, they tend to reward people who look and sound like the last person who did the job. The bias is not obvious per person. It shows up as sameness across a hundred hires.

The fix is the same idea Grede is pointing at. Define the behaviors you actually want. Then hire for skills and check the model for drift. If you cannot say what a respectful teammate does on a Tuesday afternoon, you cannot ask a system to find one.

A Practical Playbook for Leaders

If you run a team and you are adopting AI tools right now, here is where to start.

• Write down your culture rules. If it cannot be written in plain language, it cannot be automated. Vague values become convenient excuses.
• Sort AI use into two buckets. What may the system decide on its own, and what may it only suggest? Hiring, pay, and firing belong in the second bucket, almost always.
• Keep a human in the loop for anything hard. Warnings, layoffs, and serious complaints need a person, not a screen.
• Hunt for "family" language in your own documents. Search your job ads, handbook, and review templates. Replace it with concrete promises you can keep.
• Measure outcomes, not sentiment. Track turnover, overtime, pay gaps, and promotion rates. A "happy" survey score means little if the same three people quit every quarter.
• Always give workers an appeal path. If a person is affected by an automated decision, they should be able to ask a human to look again, and get an answer.
• Tell people what the tools do. Secret monitoring destroys trust faster than any policy. Transparency is cheaper than repair.
• Train managers to use AI as a rehearsal tool, not a replacement. Prepare with the machine. Deliver as yourself.

The Future: Two Paths for AI at Work

There is a fork ahead, and it is not really about technology. It is about which version of culture we copy into it.

On one path, AI becomes a tool for extraction. It tracks every keystroke, fills every gap in the schedule, and dresses up cost cutting as care. The "family" line survives, now automated, now cheerful, now impossible to argue with because there is no one to argue with.

On the other path, AI becomes a tool for fairness. It makes boundaries real by enforcing them evenly. It gives small teams the kind of backup that used to exist only in big ones. It catches the manager who forgets, the policy that gets skipped, and the pattern nobody noticed. It does not fake warmth. It just holds the line.

Grede's advice points toward the second path. Empathy without family is not a colder idea. It is a clearer one, and clarity is exactly what software needs. A feeling cannot be coded. A promise can.

The Bottom Line

The leaders who get the next decade right will not be the ones with the warmest slogans. They will be the ones who can explain, in plain words, what their people are owed, and then build systems that deliver it, day after day, without anyone having to remember.

That is the real test of leadership in an AI workplace. Not whether the machine can sound kind. Whether the company can actually be fair. Grede's point is the starting line: care about people like people, not like family. Then write it down, and let the tools help you keep your word.