I automated my job (and it made me a greater chief)

Right here’s the factor about senior management that no one warns you about: the job isn’t arduous due to any single job. It’s arduous as a result of your work lives in fifteen totally different locations and your mind is the one system connecting them.

Conferences bleed into one another. Selections are made in threads with out you. Somebody talked about your title in a planning assembly, and now there’s an motion merchandise residing in a doc you’ve by no means seen. You’ll discover out about it in two weeks when somebody casually asks for an replace. Enjoyable.

Final yr, my crew virtually missed a efficiency assessment deadline as a result of it was introduced in a channel no one was watching. One individual spent ten minutes looking out Slack and couldn’t discover it. One other discovered the date in a random, unrelated channel. I ended up posting “I’ll admit we dropped the ball on following up in Slack, in order that’s on me.” That’s the type of factor that retains taking place when your mind is the one system connecting every little thing.

I used to be spending a lot power on context-switching that I had nothing left for the pondering, connecting, and creating that my position truly requires (and that’s the work I truly like doing). However I began utilizing automations within the GitHub Copilot app, and it modified my complete workflow. Bear with me.

What automations truly are

The GitHub Copilot app is a standalone desktop app for macOS, Home windows, and Linux, constructed for working with brokers, not simply speaking to them. You’ll be able to run parallel classes throughout repositories, every by itself department and worktree. You’ll be able to see what brokers are doing in actual time by canvases, that are bidirectional work surfaces the place you and the agent function on the identical plan, terminal, or browser session. Progress is seen and steerable, not buried in chat historical past.

Automations are scheduled prompts that run towards your actual work context: your calendar, your e mail, your messages, your GitHub repos. They join by MCP servers and integrations, to allow them to see what’s taking place throughout all of the locations your work lives. They inform me what truly wants my consideration, which lets me ignore the remaining.

Consider them as brokers with a standing transient. You inform them what to care about, learn how to suppose, and when to run. Then they only… do it. Daily. With out you remembering to ask. Which is sweet, since you gained’t.

What this seems to be like

I’m a senior director at GitHub. I lead developer relations. My scope is vast, my calendar is full, and my mind works in another way than most individuals assume. I’m AuDHD, which suggests I’m good at sample recognition and deep focus, however genuinely horrible at remembering which thread I promised to observe up on three days in the past.

I didn’t got down to construct 40 automations. I used to be curious concerning the automations tab, requested the app what it might do, and it prompt issues I hadn’t considered. The primary time I set one up, I opened a chat and mentioned one thing like: “Look throughout all of my work surfaces, my calendar, my e mail, my messages, and determine the place I’m dropping balls, the place I’d need assistance, and recommend automations that may be helpful.”

It instantly prompt about six. The primary drafts weren’t excellent, and that’s okay. You refine them. You give them voice. You train them the way you suppose. As soon as I noticed what was doable, I stored going. Now I’ve about 40. (I do know. I do know.)

I’m not going to stroll by all of them. (You’re welcome.) However listed here are the classes that matter most, and a few highlights from every.

The morning transient

Daily earlier than I open something, a number of automations have already run. Assembly Prep pulls my calendar and builds context for each assembly, with totally different codecs for one-on-ones vs. giant syncs vs. exterior calls. By the point I sit down, I do know what every assembly is about and what I must carry. Pre-Assembly Entry Verify verifies I even have entry to the docs and hyperlinks referenced within the invite. No extra displaying up and realizing the agenda doc is locked. Should you’ve by no means skilled that exact panic, actually, have to be good. Day by day Triage Digest sweeps GitHub, e mail, and messages for something that wants my consideration.

The cumulative impact is that my mornings went from “frantically opening twelve tabs whereas pretending I’ve learn the agenda” to “studying a couple of summaries with espresso.” It’s a distinct life.

Staying present

I can’t be shocked by our personal launches. That’s actually the job.

Ship Decoder finds every little thing GitHub shipped within the final 24 hours and explains it to me in plain language. That is actual context I can use in conversations. Launch Radar runs weekly and surfaces upcoming launches that contact my crew’s house so I’m by no means blindsided. These two alone in all probability save me an hour a day of scrolling by channels attempting to piece collectively what occurred. I used to spend that hour. I didn’t get pleasure from that hour.

Profession structure

That is the class that shocked me most. I constructed automations that actively work on profession growth, and if that sounds bizarre, stick with me.

Day by day Wins Recap runs each night and summarizes what I truly achieved. This one issues greater than it sounds. My default mode is to examine one thing off and instantly transfer to the subsequent factor. I don’t sit with it. I don’t acknowledge it. I simply hold going. Then efficiency assessment season comes round. I’ve to articulate my influence, and I’m panic-staring at a clean doc attempting to recollect eight months of labor.

This automation retains a operating document so I don’t should. Consider it as a gratitude apply backed by actual knowledge fairly than a job listing. It counters the “what did I even do at this time?” spiral that hits hardest on the busy days. On the times when imposter syndrome is loud, I want one thing that talks again to it with details. The robotic believes in me even after I don’t. That’s oddly shifting? I don’t know. It really works.

Workforce and other people

That is the place I need to be actually sincere, as a result of I do know you may be pondering: is she automating the human elements of her job?

No. And that distinction issues to me greater than anything on this put up.

Commitments and Comply with-Up Tracker searches my very own messages for issues I mentioned I’d do and flags what I haven’t executed but. This one is humbling. And important. As a result of after I inform somebody “I’ll look into this” after which neglect, that’s a belief downside. The automation protects the belief.

The kudos I write are nonetheless mine. The noticing remains to be mine. The automation simply makes positive my mind doesn’t steal it from the individuals who deserve it.

These automations don’t exchange connection. They permit it. They offer me again the headspace to truly present up for individuals. Earlier than this method, I’d stroll into conversations distracted or operating on fumes, as a result of my mind was stuffed with operational noise. Now after I sit down in a one-on-one, I’m truly current. Once I write recognition for my crew, it’s particular and actual.

The automations deal with the scaffolding. I do the human work. That’s the deal.

Upkeep and logistics

This class covers the boring stuff that quietly eats your week in the event you let it: Dependabot PR Triage finds and merges protected dependency updates throughout my repos day by day. Dealt with. Stale Work Finder surfaces pull requests I opened and forgot, points that went quiet, branches amassing mud. (All of us have these. Don’t lie.) Journey Logistics Tracker watches for conference-related threads and consolidates logistics right into a single transient. Convention season is chaos. This helps.

What my automations appear like

Right here’s an actual one from my setup, the Stale Work Finder, so you possibly can see what these prompts truly appear like in apply:

That’s it. That’s the entire automation. You write a immediate, set a schedule, and the agent runs it in your behalf. You may get as detailed or as free as you need. The app fills in context out of your related instruments. It runs each Monday for me and the outcomes are… all the time a bit eye-opening. Nevertheless it’s higher to know.

The AuDHD half

I’ll simply say it: for me, automations are an accessibility device.

AuDHD means my govt perform and dealing reminiscence are wildly inconsistent, and the inconsistency is the toughest half to elucidate to individuals. Some days I can maintain seventeen threads in my head. Different days I neglect I’ve a gathering in 10 minutes. There isn’t any in between. The hole between these days used to scare me, as a result of my crew deserves constant management no matter what my mind is doing on any given Tuesday.

These automations slender that hole. They make me constant. They imply my crew will get the identical high quality of consideration whether or not my govt perform confirmed up at this time or not. For me, that’s the distinction between thriving and slowly burning out. And I’ve executed the burning out half. Zero stars, wouldn’t suggest.

Find out how to begin (for actual)

Should you’re enthusiastic about constructing one thing like this, right here’s what I’d say: don’t attempt to automate every little thing without delay. Begin with the one factor that causes you essentially the most friction.

For me, it was assembly prep. I stored strolling into conferences chilly as a result of prep required visiting 4 totally different instruments and synthesizing info I didn’t have bandwidth to synthesize. One automation mounted that. And as soon as I felt that aid, I stored going. And going. And going.

Might I consolidate a few of these? Most likely. I’ve about 40, and I’m positive a few of them could possibly be mixed. I favor specificity, however you could possibly simply roll a number of into one huge automation if that’s extra your type.

Right here’s the trick that labored for me: open a chat within the GitHub Copilot app and ask it to audit your work surfaces. The place are you dropping balls? The place are the repetitive patterns? What’s the factor you retain that means to do however by no means get to? Begin there.

The primary draft gained’t be excellent. That’s wonderful. You refine it in dialog. You train it your voice, your priorities, how you consider “good.” You then let it run.

Begin with one. See the way it feels.

Then construct one other. And one other. And earlier than you recognize it, you may have 40, and also you’re writing a weblog put up about it. Anyway.

The larger image

I believe we’re at an attention-grabbing second for the way individuals relate to AI at work. The early dialog about AI at work was principally about era. Make me a factor. Write me the code. The fact, not less than for me, is extra like augmentation of invisible labor. The stuff that burns you out however by no means exhibits up in your output. The meta-work no one acknowledges in efficiency opinions however everyone seems to be buried in.

Each chief I do know is overwhelmed by context. Each neurodivergent skilled I do know is spending monumental power on programs that neurotypical individuals navigate with out enthusiastic about. Automations gained’t repair organizational dysfunction or unhealthy administration or an unreasonable workload. However they may give you again sufficient headspace to truly do the work you’re right here to do.

And actually? That’s sufficient. That’s rather a lot.

And look, this can be a GitHub product. It’ll additionally run your dependency updates, triage your points, do safety sweeps throughout your repos. The developer workflows are precisely what you’d anticipate. I simply occur to make use of it for the elements of my job no one talks about.