Most management advice is about how to become a better manager. The 37signals Manager Playbook left me thinking more about how much management a company actually needs.
37signals has made the manager’s job surprisingly small. Managers are expected to remain practitioners in their field. The company says managing people “should not take much of your time,” in part because it hires people it expects to operate as “managers of one.” New employees get more attention, as do people who are struggling, but an experienced person doing good work is mostly allowed to keep doing it. The manager is there, but not constantly present.
That only works because the company has removed quite a few things that normally make managers busy. Under its boundaries for managers, employees tell managers when they are taking vacation rather than asking permission. Annual compensation is handled centrally rather than negotiated manager by manager. Profit sharing follows a formula. Serious personnel complaints go to People Ops. A manager still has authority, but there are fewer reasons to exercise it simply because the organization has routed so many administrative decisions elsewhere.
What remains is much closer to the work itself. The standards for managers start with expertise in the field and the ability to recognize good work, give useful guidance, and help people improve. Performance reviews are supposed to refer to actual work products. A manager shouldn’t drift so far into coordination and administration that the work becomes something other people do.
I find that appealing, probably because the opposite happens so easily. Someone is good at a job, gets promoted, and gradually acquires enough managerial machinery that the original job disappears. Before long, success as a manager is measured partly by how well the manager operates the machinery: meetings, approvals, staffing processes, reports, planning cycles, performance systems. Some of that is unavoidable. But it can become difficult to tell which parts are helping people do better work and which parts exist because the organization has decided that managers are where administrative work should go.
37signals seems fairly suspicious of the manager as a little center of power. Its guidance on feedback repeatedly warns against interfering just because someone works differently than you would, answering every question yourself, or turning coaching into a demonstration of your own expertise. If the work is good, the manager doesn’t need to have an opinion about every detail of how it got done.
The tradeoff is that managers have to be willing to exercise judgment when it actually counts. The company’s performance model evaluates people on skill, engagement, and coachability, and it expects problems to be addressed early and directly. Autonomy doesn’t mean pretending not to notice weak work. It means leaving capable people alone enough that, when a manager does step in, there is hopefully a real reason for it.
Clear standards make autonomy easier. If everyone has a decent understanding of what good work looks like, there is less need to supervise the process of producing it. And a manager who still knows the work can talk about the work itself rather than falling back on activity, responsiveness, process compliance, or whatever other proxies happen to be easy to see.
The same thing shows up in the playbook’s approach to one-on-ones. They aren’t supposed to be status meetings because the company already has other ways of knowing what is going on. A one-on-one can instead be about a difficult decision, work that went especially well or badly, something the employee is stuck on, or what they want to get better at. Experienced employees who are doing well may meet with a manager monthly. That sounds quite reasonable to me. I’ve never been convinced that a recurring meeting becomes valuable merely because it recurs more often.
I’m less certain about how neatly all of this travels outside 37signals. The system assumes you can hire people who are comfortable with a lot of independence, give them unusually clear expectations, and build the rest of the company so information and decisions don’t constantly have to move up and down a management chain. Plenty of organizations have more dependencies, more unevenly defined work, and more reasons for coordination. Copying the meeting cadence or the performance rubric while leaving everything else unchanged probably wouldn’t accomplish much.
The management model is not just a set of habits for managers. It is partly the result of choices the company has made elsewhere. If compensation is handled consistently, the manager doesn’t need to become the employee’s compensation negotiator. If vacation doesn’t require approval, the manager doesn’t need to approve vacation. If information is already visible, the manager doesn’t need to spend a meeting collecting status. If capable people are allowed to make decisions, the manager doesn’t need to become a queue for decisions. Each of those choices leaves a little more room for the work that is genuinely hard to turn into a process: noticing how someone is doing, helping them improve, recognizing unusually good work, and having an uncomfortable conversation before a problem gets worse.
There is even a bit of AI in the playbook. 37signals suggests using it to pressure-test hiring exercises and decisions, prepare for difficult conversations, and compare expectations across levels. I liked that it treats AI as another way to examine the evidence rather than as a substitute for making the call. The manager can get help thinking, while the judgment still belongs to a person.
I don’t know that I’d want every company to manage exactly this way. I do think there is something healthy in asking, before teaching managers to do another managerial task, whether the task needs to exist at all.
A manager who has fewer things to manage might have a little more time to notice the people and the work in front of them.