The software engineering job got automated faster than anyone predicted. Product management is next, and most PMs aren’t ready.
Emma Burrows, ex-CTO of Stripe UK, has spent 18 months doing what most PM leaders are still theorising about. She built Rezonant, a product workspace for the agentic era — to prove it. This Wednesday, she’s walking Productify readers through exactly how.
📅 Wednesday, 27th May · 45 minutes · Zoom
🌍 London 12:00 PM GMT
🌍 Amsterdam 1:00 PM CET
🌏 Mumbai 5:30 PM IST
🌎 San Francisco 5:00 AM PDT
Expect a behind-the-scenes walkthrough of a real product brain in action — the first workflow Emma automated end-to-end, what broke, and a clear starting point you can act on next week!
Here’s the writeup from Emma — straight from the person who’s actually solving the problem of agentic product work:
The job of automating software engineering is a much easier one than automating the job of a product manager. There are a lot of factors that contribute to this, but the ones that are most pertinent are:
● Software engineers historically spent a large amount of their time doing one thing (writing code)
● Code can be validated - you can tell if it at least works, and to a large extent whether it’s good
● There’s a lot of historical data for training the models
A good engineer today is running lots of coding agents simultaneously and has become an agent orchestrator.
The question I’ve been thinking about recently is which PM workflows should become agent orchestrator functions, what needs to exist to make this transition happen, and how to do it without losing taste and judgement.
Most PMs these days are using AI like engineers used Cursor a year ago - as an assistant. They feed it a bunch of stuff and it does some useful things that help speed up your day job by 20% or so. But, to really enable the kinds of productivity gains that your software compatriots have seen, you need to start working out how to hand off full end-to-end workflows to agents. The biggest unlock for software engineers was when agents became capable of long-running tasks, and I’m trying to find the same thing for product managers.
Before we get any further, I want to clarify - I hate AI slop.
I think when all of the tools we have at our disposal are used badly, an entire company can become a slop machine. In order to prevent this, but still unlock huge efficiency gains, a true PM agent system needs to:
● Learn your taste and judgement: you need to teach it how to think and act like you.
● Work to reduce your interrupts: every interrupt reduces your ability to think strategically about the things that matter that an agent cannot do.
● Automate the easy stuff that you do many times a day.
Here is the transition that I expect many PMs to go through over the next 18 months if they want to stay relevant in their jobs:
1. Use AI as an assistant to help the ‘writing’ parts of their job (like PRDs, specs, weekly updates, etc) while maintaining taste and coherency
2. Use AI in mini workflows and skills to help automate individual tasks (e.g breaking things down into user stories, synthesizing customer feedback into research reports and strategy documents). The product that we’ve got live today, Rezonant, is great for this kind of thing.
3. Use AI to repeatedly automate workflows end-to-end - with 70% eventually running completely without human intervention.
Building your product brain
Context has become a very familiar concept to anyone working with agents. Fundamentally a good product brain consists of these components:
● Inputs: Conversations with your team, your customers, your strategy, your product and its marketing, the competition
● Autonomous organization: This can slurp in the inputs and organise it for you so that you don’t spend your life moving .md files around. Otherwise you’ve just swapped one boring task for another.
Teaching it your taste and judgement
In order to help you, the system needs to understand your taste and judgement. Many people start here by trying to encode it. It kinda works and it’s not a bad starting point, but it’s quite hard to do, and not very maintainable.
Instead, I think the ideal product brain is self-learning. It should learn how to be you. Every time it gets it wrong, you should tell it why, and it should use that to maintain its own view of product principles (you can of course edit them).
That’s why the human agent interface is the final component of the product brain system. It needs to be somewhere you hang out anyway. Slack is perfect for this. You should use it for building confidence in your agent - i.e before it does any serious action, it should ask you for permission. If it gets it wrong, it writes that back to the brain as a key piece of information to learn from.
Where does this actually leave me as a PM?
I recommend using a product brain to automate the basics. If a customer comes to you with a small feature request, the product brain should be able to take it from end-to-end. Straightforward edge case question from an engineer? Same thing.
This feels like a threatening way to automate yourself out of the job. But, if you look at what has happened with software engineering, the opposite is true - the engineers that are thriving in this new ecosystem are those that have become agent orchestrators and architects.
In the same way, the goal as a PM is to use additional time available on the features that are hard to shape, that require intuition and deep thought and to make space for taste and coherency. For these kinds of projects, a bit of friction is helpful - you don’t want to ship, you want to ship with taste. That’s where prototyping, PRDs, working with design and research are not only still useful but significantly more useful than before. When everyone can ship fast, the big changes matter more.
Where this is all heading
The direction of travel is clear even if the timeline isn’t. As AI advances, we’ll start seeing ‘software factories’ - automated pipelines where defined parts of the product run without constant human sign-off. The PM’s job in that world shifts significantly; it’s less about managing execution (be it human or agentic) and more about deciding which workflows get automated and which ones stay human-guided.
The PMs who will navigate that transition well are the ones training the muscles now. Being thoughtful about where you slow down - rather than defaulting to automating everything you can - will put you in good stead as these ways of working become commonplace.
Join me to hear about how I’m building our product brain
45 minutes. No slides. We’ll cover how Rezonant’s Slack agent works, the first workflow Emma automated end-to-end — and what broke. Just for Productify readers.
📍Zoom
📅 Wednesday, 27th May · 45 minutes · Zoom
🌍 London 12:00 PM GMT
🌍 Amsterdam 1:00 PM CET
🌏 Mumbai 5:30 PM IST
🌎 San Francisco 5:00 AM PDT