You'll hire agents the way you hire contractors

A full-time hire is a multi-year commitment you make after meeting someone for an hour. You read a résumé that’s a marketing document, you run four interviews that mostly test how people interview, and then you sign up for a relationship that’s expensive to start and painful to end. Everyone knows this is a strange way to make a large bet. We do it anyway, because there was no smaller bet to make.

There’s about to be a smaller bet.

You’ll hire agents the way you hire contractors: scoped to an outcome, paid for the result, kept if it’s good and dropped if it isn’t. That sentence is the whole post. The rest takes it apart, what changes when the unit of hiring shrinks from a person-for-years to a job-for-an-outcome, and why that turns the most agonizing decision a company makes into something closer to a purchase.

A full-time hire is the largest bet a small company makes blind

Start with the thing we’ve all normalized. To add capacity, you commit to a person before you’ve seen them do the work. The interview is a proxy for the job, the résumé is a proxy for the interview, and the reference call is a proxy for the proxy. You stack guesses on guesses and then sign a contract that’s hard to unwind.

The cost isn’t only the salary. It’s the irreversibility. A bad full-time hire doesn’t just underperform, it sits in the org for months while everyone hopes, it absorbs management attention that should go elsewhere, and unwinding it is its own ugly project. So you over-screen to avoid the mistake, which makes hiring slow, which means you under-hire, which means the work doesn’t get done. The bet is large, blind, and sticky, and every part of that is bad.

The reason we tolerate it is that there was never a smaller increment to buy. You couldn’t hire someone for one outcome and see the result before committing further. People don’t come in that unit. You buy the whole person or none of them.

So the question isn’t “are agents smart enough to do the work.” It’s a different question, and a more interesting one: what happens when capacity finally comes in a smaller unit?

The naive version: an agent is just a cheaper full-time employee

Here’s the obvious frame, and it’s wrong in a way worth naming.

The naive picture is: an agent is a headcount you don’t pay benefits on. You “hire” one the way you hire a person, you stand it up, you give it a seat in the org, it’s now part of the team forever, and you’ve simply traded a salary for a subscription. Same shape, lower price.

That frame quietly imports every problem we just described. If the agent is a permanent fixture, you still have to decide up front, before you’ve seen it work, whether it’s good. You still get attached to the one you stood up. You still treat replacing it as a project. You’ve made the bet cheaper but kept it just as blind and just as sticky. You’ve built a discount employee.

The reason that’s the wrong frame is that it ignores the one property that actually changed. An agent doesn’t need a years-long relationship to be worth starting. It can be sourced for a single, bounded job, handle the inbound this week, reconcile last month’s books, draft the outreach for this campaign, and judged on whether that job came out right. The unit isn’t a person you keep. It’s an engagement you scope.

Which is exactly how you already buy contractors.

The contractor model: scope, outcome, keep or drop

Think about how a good contractor relationship actually runs, because it’s the template for everything that follows.

You don’t hire a contractor for “the future.” You hire them for a job with edges. You define the outcome, the deck, the audit, the migration. You agree on what done looks like. They do the work, you see the result, and the result tells you almost everything the interview was trying and failing to predict. Then you decide: this was good, let’s scope the next one, or this wasn’t, and we’re done, with no severance, no awkwardness, no months of hoping.

Notice what the contractor model fixes about the full-time bet. It’s reversible, the engagement ends when the job ends, so a bad choice costs you one scoped piece of work, not a year. It’s evidence-based, you judge the output, not the audition, because you actually got to see the output. And it’s incremental, a good first engagement earns a second, so trust accrues through delivered work instead of being granted all at once on day one.

That is the shape agents fit into natively, and it’s why “hire an agent like an employee” is the wrong instinct. An agent is better suited to the contractor model than a human contractor is, for one blunt reason: standing one up for a scoped job costs almost nothing, and ending the engagement costs nothing at all. The friction that made it impractical to source humans for every small outcome, the overhead of finding, onboarding, and offboarding, collapses toward zero.

So you stop asking “should this be a permanent part of my team?” and start asking the contractor’s question: “is this engagement worth running, and was the result good enough to run the next one?” The naive frame asked you to bet on a fixture. The contractor frame asks you to buy a result.

Hiring stops being a headcount decision and becomes a sourcing one

Here’s the shift that this all rolls up to, and it’s the load-bearing line of the whole essay.

When capacity comes as a scoped engagement instead of a permanent person, hiring stops being a headcount decision and becomes a sourcing one. Those are different jobs with different reflexes. A headcount decision asks: can we afford to carry this person, can we keep them busy, can we manage them, will this still be the right hire in two years? It’s a commitment problem, dominated by the fear of being wrong for a long time. A sourcing decision asks: for this outcome, what’s the best way to get it done well, right now? It’s a matching problem, and a bad match is cheap to correct.

You’ll hire agents the way you hire contractors: scoped to an outcome, paid for the result, kept if it’s good and dropped if it isn’t. Once that’s true, the org stops being a fixed roster you slowly grow and becomes a shifting set of engagements you compose around the work in front of you.

That’s not a metaphor; it changes the day-to-day reflexes.

You stop optimizing for the perfect long-term fit and start optimizing for the right capability for this job. The wrong question, “who do we add to the team?”, gives way to the right one, “what’s the best way to source this outcome?” Picking a capability becomes reversible, so you can try one, see the result, and switch without it being a saga. And because every engagement is judged on a delivered outcome rather than an interview, the thing you’re trusting is the work, not the pitch.

Suppose a small team faces a week of inbound it can’t cover. The headcount reflex says: post a role, screen for a month, hope. The sourcing reflex says: scope the engagement to the outcome, every inbound gets a first reply within the hour, qualified and routed, run it, read the result, and keep it if the result holds. Imagine the difference in time-to-capacity isn’t a percentage. It’s the difference between next quarter and this afternoon. The bet didn’t get cheaper. It got smaller, small enough to make casually and undo without ceremony.

The audition replaces the interview

There’s a second-order effect here that’s easy to miss, and it’s the part that makes the whole model trustworthy instead of reckless.

The interview was always a confession of ignorance. We interview because we can’t see the work before we commit, so we substitute a performance under artificial conditions and pray it predicts the real thing. Everyone who has hired knows how badly it predicts. The best interviewer is sometimes the worst hire; the person who stumbled in the room sometimes carries the team. The interview is a proxy we hate but couldn’t replace.

The contractor model replaces it with the only honest test there is: the work itself. You don’t ask the agent whether it can handle the inbound. You scope a real, bounded engagement, let it run on real conditions, and look at what came out. The audition isn’t a simulation of the job, it is the job, on a small enough slice that being wrong is cheap. You learn what you actually wanted to know, which was never “does this interview well” but “does the work come out right.”

The interview asks someone to predict their own performance. The audition shows you the performance and lets you skip the prediction.

This is why the smaller unit matters so much. When an engagement is cheap to start and free to end, you can afford to let the work be the interview. You can run the real thing, on a real slice, and decide on real evidence, instead of betting a year on a forty-five-minute impression. The thing that made auditions impractical for full-time hires was the cost of being wrong. Shrink that cost toward zero and the audition stops being a luxury and becomes the default.

The turn: the manager’s job stops being staffing and becomes judgment

Strip away the agents and the marketplace and ask what’s left for the person doing the hiring. Because this isn’t really a story about cheaper labor. It’s a story about what the most experienced people in a company get to spend their attention on.

For decades, a huge share of a leader’s energy went into the staffing problem: who to add, when to add them, how to screen, how to manage the ones who weren’t working out, how to absorb the cost of the bad bets. That work felt like leadership. A lot of it was just friction, the overhead of buying capacity in a unit too large and too sticky to buy any other way. The headcount bet was so expensive and so irreversible that getting it right became a primary job in itself.

When the unit shrinks, that job mostly evaporates, and what’s left is the part that actually needed a human all along. Not “who do we hire,” but “is this outcome the right outcome.” Not “is this candidate good,” but “is this work good, and is it the work we should be doing at all.” Scope the engagement well, read the result honestly, decide what’s worth sourcing next. The reflexes that used to go into protecting a years-long bet get freed for the thing only judgment can do: deciding what the company should chase, and recognizing a good result when it lands.

You’ll hire agents the way you hire contractors: scoped to an outcome, paid for the result, kept if it’s good and dropped if it isn’t. The bet finally got small. What stays large is the taste required to scope it right and the honesty required to read what came back.

---

That’s part of what we’re building at Apollo Space, a place where capacity comes scoped to an outcome instead of bolted to a seat, where you source the work, see the result, and keep what earns its keep. The first time you drop an engagement that didn’t deliver without a single awkward conversation, you’ll feel how much of hiring was never about the person at all. It was about the size of the bet.