How AI is changing the day-to-day of project managers in software teams

I’ve been thinking a lot lately about how much time project managers spend on work that isn’t really the core of their job. In software teams especially, a surprising amount of energy goes into things like labeling issues, assigning them to the right person, spotting duplicates, writing weekly reports and taking meeting notes. Necessary tasks, of course, but not exactly the ones where a PM creates the most value.

What’s interesting is that we’ve finally reached a point where a lot of this can be delegated to AI. And not in a futuristic “one day” sense. This is happening right now.

Linear, the task management tool we use, can already handle issue triage fairly well. It's able to tag issues, suggest the right assignee and even detect when something looks like a duplicate. Things that used to take chunks of time now feel almost automatic. Moreover, we’ve been trying out ChatGPT’s browser to help us draft weekly reports. The workflow is simple: we open the issue board through the browser, filter the items that had activity during the past week and let the model review each issue in detail. It reads the comments, understands what was done and then uses Atlas to turn all that information into a clean, well-written report in our own style. Something that used to take me an hour now takes just a few minutes.

We use this prompt for every issue card you see on the current board view:

1. Open the issue.

2. Capture basic info

• Issue title

• Direct URL to the issue

• Current status/column on the board

3. Check the change history

• Open the issue’s activity panel.

• Identify all changes made in the last 7 days, including but not limited to:

-Status transitions
-Description edits
-Labels, priority, milestone changes
-Assignee changes
-Due date or estimate changes

• Summarize these changes briefly in your own words.

4. Read new comments

• Scan comments from the last 7 days.

• Extract a short qualitative summary of what those comments indicate (progress, blockers, decisions, test results, etc).

5. Determine this week’s state

• Based on status + recent activity + comments, classify the issue into one of:

-Completed
-In Progress
-Pending / Not Started

6. Write a qualitative weekly summary

• If Completed:

-Explain what was delivered this week.

-Explicitly state delivery phase:

-“Pending production deploy”, or
-“In testing/QA”, or
-“Fully delivered and deployed to production”.

• If In Progress:

-Explain what advanced this week.

-List what is still pending or the next concrete steps.

-Note blockers if any.

• If Pending / Not Started:

-If there was no activity this week, say so clearly.

-Provide a 1–2 sentence description of what the issue is about.

-State that work has not started yet.

Style and constraints

• Be concise but informative.

• Do not copy raw logs; summarize in clear English.

• Use bullet points only when it improves clarity; otherwise short paragraphs are fine.

• Every issue on the board must appear in exactly one of the three sections.

Now start from the currently open board and proceed issue by issue until all visible issues have been processed. Then generate the final report.

And the same goes for meeting notes. Anyone who’s been in the situation of trying to contribute to a conversation while also writing everything down, knows how draining it can be. Having AI generate reliable notes, highlight decisions and create action points is a huge relief. It makes it easier to prepare future calls and reduces the chances of misunderstandings.

So the real question is: if AI takes care of the administrative load, what does the project manager focus on?

My view is that this shift lets PMs spend their time where it truly matters. Instead of wrestling with bureaucracy, we can deepen our understanding of the product, think through edge cases, clarify functionality, help the team stay unblocked and improve the communication flow. We can spend more time spotting risks early, noticing patterns and understanding how the team is performing without rushing through the details. And that’s the part of the job that actually moves a project forward.

At MarsBased, we’ve always seen the PM role as much more than a coordinator. It’s someone who understands the project in depth, both functionally and technically. Someone who can take decisions, guide engineers, translate needs between stakeholders and ensure the product makes sense end-to-end. That’s why our PMs also do QA. It’s not about testing for the sake of testing, it’s about knowing the product well enough to anticipate problems before they land in production.

And if your team could benefit from this kind of senior, AI-enhanced project management, we can help. At MarsBased, we combine deep technical expertise with a structured, calm and reliable way of working to help companies ship better software with fewer headaches.

AI doesn’t replace any of this. What it does is remove the noise. It clears the space so the PM can play a more strategic and more human role: organizing, thinking, deciding, connecting the dots. And honestly, that feels like a good evolution. The less time we spend pushing digital paperwork around, the more time we have to actually build better products with our teams.