> Some manager up the hierarchy hires the executive assistant.
My post assumed team autonomy, where each team has an entirely separate hiring pipeline. Think “startup that’s recently been acquired by a big company, but hasn’t yet been absorbed into it.” I think Amazon’s AWS service teams are also like this. Tiny little independent business units.
But really, you don’t need that; you could still have a mostly-hierarchical organizational structure, but just take hiring/firing away from the managers at every level. with no managers making hiring decisions at any level. You’d give all hiring/firing power, instead, to teams; or rather, to each team’s hiring subcommittee (which would be a temporary thing, elected anew from the team members each time hiring/firing must be done.)
You can still have traditional managers in such an org, but their main purpose at that point (if the EA duties are being played by an actual EA) would be to serve as the contact point for a team — the “API” that the rest of the org uses to access the team members and their labor. So, essentially, a talent agent.
> edit: And since the EA isn't actually accountable for productivity
Sure they are. Just because something may fail for reasons entirely out of your control, doesn’t mean you’re not usually, partially in control. A wilderness guide can still be held accountable for the safety of those in their charge, even if sometimes there’s just suddenly an angry lion.
The point of the EA—like the point of an effective manager—would be to (attempt to) be a multiplier for the team’s aggregate productivity, usually by subtly enabling knowledge-sharing and communication, while deflecting outward pressure.
As such, a measure of the EA’s effectiveness, is the same as a measure the team’s aggregate productivity, against a pre-existing baseline measurement of the team’s aggregate productivity with no EA/manager.
I would guess that it’s very likely that the average EA/manager has neutral or negative impact on aggregate team productivity. Which makes sense; an EA/manager needs to have (much) higher coordination skills than the average human being, in order for their inserting-themselves into coordination problems to be worth the overhead it introduces.
> Technically, they only need 50% of the people at best.
If you like — and this works especially well when done remotely — you can set up your EA/manager as a (team-specific, unlimited-scope) helpdesk, where team members “file tickets” with the EA to do things for them; and where the EA/manager can also file tickets themselves “on behalf of” someone, when they notice something going subtly wrong and are trying to help without being explicitly asked to do so.
Probably you could do this in a streamlined, ChatOps way, where any new team-workspace message thread with the EA/manager is automatically a new ticket.
Then, you can throw industry-standard IT-helpdesk ticket-resolution performance metrics at the tickets the EA/manager is creating+resolving.
This won’t tell you anything about aggregate productivity, but it will tell you — in a very legible, statistical way — whether the EA/manager is preferentially solving problems for only a subset of team members. (Don’t hook anything automated up to that fact; you need manual review, because it might be the team member that’s giving the EA stupid asks that they’re rightfully ignoring.)