AI will retire the micromanager

The business world has a micromanagement problem. If you’ve worked for even a few years, odds are good you’ve run into someone (a supervisor, a project manager, a “coach”) whose interactive style is some unholy mix of intrusive, pedantic, doubting, and/or controlling.

This is perhaps especially the case in software dev, where so much of what we do feels (and often is!) hidden or mysterious to the broader business. Enter the project manager, the scrum master, the technical program manager, the engineering lead… some person given the unenviable assignment of trying to (a) understand the state of work, (b) verify that state, (c) capture it in a reliable, readable way, and (d) attest to its accuracy.

Small wonder that the people tasked with figuring all this out may sometimes come off as a little… nosy.

Nobody likes a micromanager, and nobody wants to be one. Well then, good news all around. AI will put micromanagement to bed.

Graduating from sentiment to data

In starting Socratic we first set down, in ink, on a whiteboard, a list of the questions we never wanted to have to ask someone again. Among them:

• “So… when do we think this work will finish?”

• “Does anyone on the team have some spare capacity?”

• “What work is at risk, and why?”

• “Who needs help?”

• “How long would it take us to deliver this new thing?”

Traditionally, these questions involved a chat thread, a conversation, a meeting—often all three, often more than one—to come up with some semblance of an answer. Why? Because what we were mostly doing was rounding up everyone’s opinions and feelings about the question, and then trying to consolidate those opinions and feelings into some figure or name that everyone would co-sign.

Tedious? Yes. Accurate? Sometimes. Time consuming? Always.

It’s in this digging and sorting and sifting of group opinion that the micromanager’s worst tendencies are activated. Opinions multiply. Patience runs short. An answer is needed. Rinse, repeat.

The trick is to move from sentiment (”I think…” “I heard…” “My guess is…”) to data. The answers we need already exist. It just takes AI to find them.

Information fracking

The investor Tomasz Tunguz has described LLMs as a means of “information fracking.” Take a massive document, and let AI draw out its salient points.

With Socratic, what we’re fracking is the work activity data in build systems like Jira and Git. And this fracking is conducted not with an LLM, but with the “other” kind of AI—the machine learning kind.

For example. The average Jira project is riddled with signal data about how long work takes to finish—actuals for every person and type of work, layered like geological strata, starting from right now, today, and going back to the beginning of time. (Or at least, the beginning of the Jira instance.) That kind of dataset is just a feast for machine learning.

This means vastly more reliable forecasts for how long a given epic, a given release, a given initiative will take to finish. (In the words of one customer, “I don't know how you do it, but that forecast feature is pretty accurate!”) And it also means never again having to sit and debate what a story point means, or be asked “When’s it going to be done?” Those kinds of meetings, along with their micromanaging masters, get put out to pasture.

Let’s get carried away

As a founder, you’re contractually obligated to get carried away by the promise of whatever it is your company does. So allow me a little rope here…

Is it insane to say this sort of AI might help to create a more pleasant workplace? We’re talking about dissolving some of that claustrophobic experience that comes when working with micromanagers. Let teams alone to work, and let the AI sort when the work’s going to be done. Project managers can focus on supporting the team, managing stakeholders and scope, getting people unblocked, etc. True management, not the micro- kind.

If a team feels supported, and trusted, and unbothered, then they can work in a more comfortable and confident way. Too often I think, building a good team culture gets pushed to the side in favor of more-meetings-to-answer-the-same-sets-of-recurring-questions. When an AI can answer those same questions, more accurately, and with no meeting time, this leaves managers the ability to spend more time creating the culture they’re after.

But if I’m wrong about this utopia? If all we’re really talking about is using AI to claw back a few hours in the week, and to give better answers than I’m otherwise likely to get? Hell, I’ll take it.