AI in management: what works, what doesn't
For two years, "AI will transform management" has become a recurring headline. Conferences talk about it, SaaS vendors use it as their pitch, blog posts multiply. The result: many EMs have tried tools, been disappointed or excited for the wrong reasons, and come away with no clear conviction about what's actually worth it.
Here's an honest assessment — what genuinely adds value, what's noise, and how to think about using AI in your day-to-day as a manager.
The AI promise for EMs
The central problem for an engineering manager is scattered information. You have notes in ten different places, technical activity in GitHub, meetings in your calendar, tickets in Jira or Linear, commitments made in 1:1s you don't have time to re-read. All of this builds a context you have to constantly reconstruct from memory before every important meeting.
That's exactly where AI can help. Not to replace your judgment, not to automate your decisions — to aggregate, synthesize, and give you the right context at the right moment.
What actually works
Meeting preparation
Generating a brief before a 1:1 — "here's what happened with this person this week, here are the pending commitments, here are the signals to note" — is a concrete application that saves time and improves the quality of conversations. No magic: the AI synthesizes what already exists in your tools. But that synthesis, done manually, would take you 15 minutes. Done automatically, it takes zero.
Structured information extraction
After a meeting or a 1:1 note, automatically extracting actions, commitments, and mentioned risks is something AI does very well. The raw text stays as-is, but structure emerges from it automatically. You no longer have to re-read your notes to figure out "what was decided again?"
Activity signals
Analyzing Git activity patterns to extract insights — "this person pushed a lot of small fixes late at night this week, which is unusual" — is a solid application. It's not surveillance, it's context-reading. What you were already doing by looking at PRs, but at scale and with more continuity.
Conversational assistant
"What are Lucas's goals for this quarter?" or "What feedback did I give Sarah last month?" — asking natural-language questions about your managerial memory genuinely changes how you access information. No more digging through nested docs.
What doesn't work (yet)
Making decisions for you
AI doesn't know whether you should promote someone. It doesn't know whether the conflict between two team members comes from underlying tension or passing stress. Management decisions involve human context, relational nuance, and situational reading that data doesn't fully capture. Be wary of tools that claim otherwise.
Replacing active listening
Having the best possible brief before a 1:1 doesn't replace being truly present during the conversation. AI can give you the right starting points — the direction remains human. If you're using the brief to "check boxes" rather than open a conversation, you're missing the point.
Working without data
AI is only as good as the data it's given. If you don't take notes, if your integrations aren't connected, if commitments aren't captured somewhere — the brief will be empty or generic. AI amplifies an existing practice; it doesn't create one from scratch.
The right framework to leverage it
The right question isn't "how do I use AI in my management?" but "which repetitive tasks take up my time and degrade the quality of my interactions?" If the answer is "remembering what was said, finding pending commitments, having context before a meeting" — then AI has something to offer.
The right use is augmentative: AI takes care of memory and synthesis so you can focus on what matters — the relationship, the judgment, the direction.
That's exactly what Moston is trying to do. Not to automate management, but to free you from the cognitive load that prevents you from being truly present for your team.