An Engineer with AI
In the early part of 2026 I gave a talk at a local meetup about how I am using AI in the workplace. That was 9 months ago and it's a bit crazy the trajectory we are on and how much AI I leverage in the day to day.
If I started with the morning of work I normally roll into the office with anywhere from 80 to 200 unread emails. Thankfully most of those emails are just the GitHub notification emails from changes from our Australia office from their day and our night. I can skim subject lines and quickly see the files changed directly from the email so I know if it's anything I want to go visit (CI/CD changes, package changes, Docker changes, etc). I'm sure AI could help here, but with keyboard shortcuts 10 minutes every morning isn't the end of the world.
What remains after that is mostly client emails that Google (Gemini) has summarized on top. In the beginning I really hated that AI was taking up vertical screen space for a summary that I could already glean from the subject line. However, there are a few email chains each week that float over 100 emails in the thread and since those responses might come days apart that Gemini summary is very helpful. It isn't always perfectly accurate, but it's just enough context to allow my brain to remember what the last few responses to the thread have been without having to re-read.
A few of those emails are client requests, support requests or things to dig into. With GitHub and items I feel pretty confident an AI can do unguided I can dispatch a Copilot agent to go fix/change something in the system. The beauty of this process is I'm not limited to a single occurrence. I can dispatch as many as I need across as many projects as I need so a Monday morning might reach double digits of agents running at once.
Now you have to be realistic here that I've only moved a bottleneck. If we once had a bottleneck in understanding a request and getting it front of an engineer to code, now I've just loaded up engineers with a ton of code to review. Thankfully in most cases if you balance an agentic workflow with tasks your gut says an AI can do without human support the ending result is probably a small pull request.
So after my email queue is done, I might have dispatched a ton of AI agents doing small requests that I'll continue to follow up on until I'm happy with them or I determine AI cannot help us. Which leads me to either taking an item across the finish line myself, closing it out entirely from AI's turn or pass to a queue for a coworker.
The work that remains at this point is usually just questions or tasks that have no code associated with them. I have found extreme support on either OpenCode or Claude Code for these tasks.
At one point I had to do a bunch of stuff with Android applications on a device. I knew what I had to do, but I gave the instruction set to OpenCode and it communicated with my connected device and knocked the task out. I was even happy it asked a few questions along the way thus knowing it would be done exactly as I wanted in a fraction of the time. This is a dangerous amount of power, because I wonder if my skills are atrophying by these actions.
It's an interesting slippery slope I feel is developing. I just don't feel I have the right way to describe it yet. Engineers have existed for a reason, but I've met a few friends who've lost their job to what they believe is an AI revolution. It's tough to know what to believe - an AI makes me more productive and quicker than I've ever been. Though some believe all they need is AI instead of a human and some believe they are building the solutions that will remove their own job.
So when I sit there invoking Claude Code or OpenCode, while babysitting some GitHub Copilot PRs while obtaining emails from meetings summarized by an AI it seems like my job has massively changed from what it once was. I may be an orchestrator of agents, I just have to be careful that I still understand what they are doing. Since at any moment the AI's could go away and I want to remain sure that means nothing to me except a decrease in productivity.