A while back I wrote this piece ‘Should CPOs write user stories’ – in essence what I was getting at was not really whether CPOs should write user stories but part of a larger question – how far should CPOs, product managers or leaders get their hands dirty? How much coding knowledge should they have? Is SQL a must have or can you lead by delegating even basic tech tasks whilst leading product teams?
AI Agents might be the one thing that product practitioners need to go beyond understanding and move into curation and execution. An analogy that I can think of is when product designers (Who might at first have been visual designers or web designers many years ago) started to incorporate CSS and other coding into their toolkit. Indeed now they are taking on more and more deeper ‘dev’ capabilities.
Product managers are constantly juggling priorities, analysing user feedback, tracking metrics, and making decisions with incomplete information. Typically they might rely on data scientists to work out how to automate much of this and providing dashboards in third party platforms to make sense of it all.
But AI agents – intelligent systems that can act autonomously to complete tasks, make decisions and interact with tools are here to shake things up yet again.
Unlike simple chatbots that just respond to questions, AI agents can take action, use multiple tools, and work through complex workflows without constant human intervention.
What Makes an AI Agent Different?
Think of the difference between asking someone for directions versus hiring a personal assistant to plan your entire trip. A traditional AI tool (like ChatGPT in basic mode) is like asking for directions—you get an answer, but you still have to do all the work. An AI agent is like the personal assistant—you tell it your goal, and it figures out the steps, uses the right tools, and gets the job done.
AI agents typically have four key capabilities:
- Autonomy: They can work independently once given a goal
- Tool Usage: They can interact with APIs, databases, and software tools
- Decision Making: They can evaluate options and choose the best course of action
- Learning: They can adapt their approach based on results and feedback
Why Product Teams Should Care About AI Agents
We’re all now familiar of how ChatGPT can carry out repetitive tasks. AI agents excel at exactly these kinds of workflows and take what we’ve been used to on one level with ChatGPT, Claude etc to another level!
User Research Analysis: Instead of manually categorising hundreds of user feedback submissions, an agent can analyse themes, extract key insights, and even draft initial product requirements based on common user needs.
Competitive Intelligence: An agent can continuously monitor competitor websites, app store updates, and social media mentions, then summarise changes and flag significant developments.
Data Pipeline Management: Rather than manually pulling metrics from different dashboards and creating weekly reports, an agent can gather data, perform analysis, and generate insights automatically.
Customer Support Escalation: An agent can analyse support tickets, identify product-related issues, and create JIRA tickets with proper categorisation and priority levels.
A/B Test Management: From setting up experiments to monitoring results and flagging statistically significant outcomes, agents can handle much of the mechanical work around experimentation.
The Real Power: Compound Tasks
Where AI agents really shine is in handling compound tasks—workflows that require multiple steps, decisions, and tool interactions. For example:
“Monitor our top 5 competitors’ pricing pages weekly, detect any changes, analyse the impact on our competitive positioning, and create a Slack alert with recommendations if any competitor drops prices by more than 10%.”
This single instruction encompasses web scraping, data analysis, business logic evaluation, and communication—tasks that would normally require hours of manual work or complex automation scripts. These are things that previously I’d have to rely on my devs or data scientists to carry out.
But should product managers be tasked to building the AI agents or simply have an understanding of them?
The Strategic Advantage
For product teams, AI agents represent more than just automation—they’re a competitive advantage. While other teams are still manually processing data and insights, your team can focus on high-level strategy, creative problem-solving, and building relationships with users and stakeholders.
The key is starting small, learning the patterns, and gradually expanding your agent capabilities.
As a product person I’m always curious and I do think that sometimes it’s great to go down these learning rabbit holes – you’ll get stuck, spend hours in Terminal trying to work out directory issues in order to get the code your GPT provided for your agent to work but it is a valuable journey!
Product managers need not have to be able to create agents from scratch but they certainly need to understand the power and rapidly evolving space that we’re in! There’s also no harm in trying to build a basic AI agent (Something I’m experimenting with) to look over customer feedback for example and provide recommendations. Sure, the inputs and outputs are not necessarily ‘founder friendly’ (more dev friendly!) at the moment but roll your sleeves up and have a go!
I found this video which I found really informative that gives you some context over how now UI is being simplified to gain access to deep functionality.