识别并保留“不可替代者”：角色重构与绩效评估方法

There’s one thing you will look back on regret: letting one of those truly Irreplaceable folks on your team … leave.

Mainly it happens as you scale. Things chance. You need larger teams, different processes. And you bring in new executives that maybe don’t seem them as quite so … Irreplaceable.

But these are the ones that truly know how it all works. Why it works. And they’re often the best ones to help you figure out how the next product, territory, geo, key partnership, etc. can work.

Ask your boss today what 1 thing you can take off her plate. And just … go do it. This may backfire. It may not be appreciated. Or … it just might bond you for life. Because no one does this.

Across a whole career you end up with a very short list of people like that. Five names. Maybe ten if you’ve been at it a long time. People you would hire again tomorrow, sight unseen, into whatever job you happen to have open.

The Irreplaceables.

And almost every B2B founder at some point lets one go because the org chart didn’t fit anymore.

Run a small firm. The associates who did this unprompted are the ones I still pick up the phone for a decade later. The ones who asked 'is this billable?' first — I can't remember their names. Did anyone ever do it for you?

— Jing (@Jinglexvector) September 4, 2026

The Test Is Whether You’d Find Budget in 48 Hours Before Knowing the Job

A top performer is someone you’d rehire into their role. An Irreplaceable is someone you’d rehire into a role that doesn’t exist yet.

Run the test in your head. That person emails you tomorrow and says they’re free in 30 days. Do you find budget before you’ve worked out what they’d actually do?

Most very good employees fail that test, and that’s fine.

What you’re buying with an Irreplaceable isn’t a skill set. Skill sets go stale in about 18 months right now. You’re buying judgment, ownership, and near-zero management overhead, and those travel across every function in the company.

Three Traits, and Performance Reviews Only Catch Part of One

• They take work off your plate without being asked, and they finish it. All the way to done, without flagging it in Slack first and waiting for permission. The tweet describes exactly one behavior and this is it.
• Their output doesn’t change when nobody is watching. This trait decides whether you can actually run a small company. Some people do great work under supervision and 60% of that work without it. A Keeper produces the same work either way, so the marginal cost of having them on the team is close to zero.
• They bring you bad news early. Everybody delivers bad news eventually, once it’s already visible on the dashboard. Keepers tell you in week one when it’s still cheap to fix, and they tell you even when it makes them look bad.

Performance reviews measure a version of the first one. They don’t measure the other two at all, which is why companies lose Keepers whose last rating said “exceeds expectations.”

The Reasons You Lose Them All Sound Rational in a Board Meeting

• The role got outgrown. The person who took you from $2M to $20M gets layered under a VP who has “done $50M,” and they’re gone in 90 days. Sometimes you do need the VP. You almost never need to lose the person who got you there. Founders collapse those two decisions into one all the time.
• Reorg math. Headcount plans are made of boxes. A Keeper is usually doing four jobs that live in three boxes, so in the redesign theirs gets merged into someone else’s and their scope quietly shrinks. It’s rarely a decision anyone remembers making.
• Comp bands. HR says the band tops out at X. The Keeper is worth 3X and knows it. Bands are there to stop bad decisions at scale, and this is the one case where you override yours.
• They got bored and you didn’t notice. Keepers rarely complain first. They start taking the recruiter calls they used to delete. By the time they tell you, they’re 80% out the door.
• An acquisition. I watched this up close after Adobe bought EchoSign. Big-company org charts have no box for the person who does four things at once. The free edition that drove a huge share of our signups was killed within weeks of my leaving. The people who knew why it mattered were still there. They just no longer sat anywhere the decision passed through.

Inventing a Role Is Cheaper Than a Search and Roughly 10x Faster

Price the alternative. An outside VP search runs four to six months, real recruiter fees, and in my experience close to half of VP hires don’t make it through the first 12 months. Then you start over.

Against that, a role you invent for someone already inside: they’re ramped, you’ve seen their judgment under pressure, and there’s no cultural risk. Even if the role turns out to be 60% right, you’re ahead of the search you didn’t have to run.

Invented roles fail when they’re executed lazily, and it looks the same every time. Vague new title, no scoreboard, “special projects,” and six months later the person feels sidelined and leaves anyway. You created the exact outcome you were trying to avoid.

What makes it work:

• Attach a number and a date. One paragraph, written down, with a metric they own and when it gets measured. Invented does not mean undefined.
• Give them the title they’d get externally, not the title that fits the band. A title that’s a downgrade in market terms makes them more recruitable, not less.
• Give them real budget or a real P&L line. Authority with no resources attached reads as a demotion, and they’ll read it that way within a month.
• Tell the team why, out loud. Unexplained, it looks like favoritism. Explained, it tells everyone else what behavior gets rewarded here.
• Revisit at two quarters. Invented roles should either harden into a real function or get re-invented. Letting one drift for a year is how you end up with a Chief of Staff with no scoreboard.

At 3 Humans and 21+ Agents, Every Human Has to Be a Keeper

We run SaaStr AI today with three humans and 20+ AI agents in production, and that moves this in two directions at once.

The cost of a non-Keeper goes way up. When one person’s leverage is 10x what it used to be, so is the damage from someone who has to be checked, managed, and re-directed. There’s no bench to absorb it.

The value of an Irreplaceable goes up more. Capacity stopped being the hard part. Agents produce enormous volume and report that all of it worked. They don’t walk in and tell you the number is wrong, the customer is unhappy, or that we shipped something we shouldn’t have. That still comes from a person, or it doesn’t come at all.

The other effect matters more than people expect: roles now change every quarter. The job someone was hired into in January genuinely doesn’t exist in June, because half of it got automated and the rest turned into something else. Hire against a job spec and you’ll churn your best people every two quarters through no fault of theirs. Keep the judgment and re-cut the job, and you compound.

If You’re Not the Founder, Just Go Do the Thing in the Tweet

Ask what to take off her plate. Then do it, all the way to done, without the status update halfway through asking for permission. It may backfire. It may not get acknowledged for a year.

Asking “is this billable?” first is completely rational, and it also caps your career at exactly the value of the hours you bill. The people who get the call a decade later never asked.

Most of the time the payoff isn’t the promotion or the raise. It’s that someone who now runs something meaningful picks up the phone when your name shows up, once, at the moment you need it most. No raise you could have negotiated is worth as much.

Make the List. It Should Take Five Minutes.

Write down the names. Everyone you’d hire tomorrow into any role, before knowing the role. If it takes longer than five minutes, or runs past ten names, you’re writing some other list.

Then two things.