"AI will make product managers obsolete."
"Every PM needs to become an AI expert."
"If you're not using AI, you're falling behind."
These are just some of the headlines flooding our LinkedIn feeds and industry newsletters. As a product manager, you've likely encountered similar proclamations, each one adding another layer of pressure to an already demanding role. Yet when I speak with product teams, I often hear a different story – one of confusion, skepticism, and sometimes even guilt about not embracing AI more fully in their daily work.
The reality behind the hype
In a study conducted by MIT of over 1,000 scientists using AI tools, researchers found something unexpected: while productivity increased significantly, job satisfaction actually decreased, with 82% of scientists reporting reduced satisfaction with their work.
This paradox of increased productivity coupled with decreased satisfaction tells us something important about the human side of AI adoption.
"I couldn't help feeling that much of my education is now worthless," remarked one scientist.
This sentiment likely resonates with many product managers watching AI developments with a mix of fascination and unease.
Despite these challenges, the potential impact of AI on how we work is too significant to ignore.
Our latest research from Atlassian’s Teamwork Lab shows that when it comes to getting the most out of working with AI, mindset matters far more than adoption – and is key to maturity.
The results are clear: the future of work is human-AI collaboration. Our data proves that the most strategic AI collaborators:
- Leverage AI to get 2x the ROI on their efforts
- Save 105 minutes daily – equal to an extra workday each week
- Are 1.5x more likely to reinvest time saved into learning new skills
- Are 1.8x more likely to be viewed as innovative teammates
We've identified patterns in how AI tools can both enhance and potentially hinder product management work. In this article we explore a model focused on maintaining human judgment and creativity while leveraging AI's analytical capabilities - a balance that research suggests is crucial for successful AI adoption.
To have any chance of success in the new AI-enabled work environment, we suggest taking the following steps:
- understand how AI works and what are its limitations,
- collaborate with AI teammates and gradually build a muscle for this type of interaction,
- and eventually anticipate where AI can add most value to your work.
Understanding your AI teammates
Before we dive into building habits with AI, we first need to understand how they operate.
Imagine you're at a party where everyone is having multiple conversations simultaneously. You're trying to follow one particular conversation, but you need to pay attention to various elements: the words being spoken, the tone of voice, previous comments, and the overall context. This is similar to how transformer models – the backbone of modern AI – process information.
Transformer models, introduced in 2017, revolutionized AI by being able to pay attention to multiple parts of input simultaneously, much like how you can process different aspects of a conversation at once. This ability to handle context and relationships between different pieces of information is what makes modern AI tools so powerful.
Large Language Models (LLMs) build on this foundation. Think of them as having read virtually everything on the internet – books, articles, code, conversations. They've learned patterns in this vast amount of data, allowing them to understand context and nuance, generate human-like text, recognize patterns and relationships, and adapt their responses based on the conversation.
In this article, we're focusing specifically on AI agents – one of the most common interfaces for interacting with LLMs. These agents act as conversational interfaces that allow us to interact with LLMs in a natural, dialogue-based way. While LLMs can be integrated into products and workflows in many ways (like code completion, content generation, or automated analysis), agents represent a particular paradigm where the AI takes on a more collaborative, assistant-like role. This is the interface that most product managers will interact with directly in their day-to-day work, whether through general-purpose tools like ChatGPT and Claude, or specialized workplace agents like Rovo.
Capabilities and limitations
AI agents excel at pattern recognition, spotting trends and connections in large amounts of data that might escape human notice. They're remarkably good at processing language, understanding and generating human-like text across various styles and formats. Their ability to maintain context across long conversations and consider multiple aspects of a problem simultaneously makes them particularly valuable for complex tasks like product management.
However, understanding their limitations is crucial. Despite their confident tone, they can present incorrect information without hesitation. They don't truly understand cause and effect, instead relying on pattern matching rather than genuine comprehension. Their knowledge is often limited to their training data, meaning they can't help with real-time market changes or emerging trends, specially when they do not have access to the internet and search capabilities.
Understanding these strengths and limitations helps us approach AI as a powerful complement to our work rather than a replacement for our core skills. With this foundation, let's explore how to build effective working relationships with AI tools.
Understanding the nuanced impact
Consider this: for fifteen years, Google was the unquestioned go-to tool for online searches. It was muscle memory - need information? Open a new tab, type "google.com ". This deeply ingrained habit seemed unshakeable. Yet in just the past year, that decades-old behavior has been completely transformed. First came Arc Search with “Browse for me”, then a rapid progression through ChatGPT, Claude, Perplexity, and now DeepSeek R1 for work-related searches.
This transformation illustrates something crucial about AI adoption: when AI tools provide a genuinely better experience - faster, more efficient, more insightful - they can reshape even our most ingrained behaviors in remarkably short timeframes. We're not changing our tools because we're told to, or because it's trendy, but because these new AI-powered solutions are fundamentally better at meeting our needs.
Instead of seeing AI as either a threat or a silver bullet, we need to view it as a new type of teammate. Like any good colleague, AI has specific strengths and limitations. The key is learning how to collaborate effectively.
Collaborating with your AI teammates
The current landscape of AI usage reveals an important pattern: most people are using general-purpose AI agents trained on general-purpose data to ask specific questions about their work. This mismatch is often the source of disappointment and frustration.
The future of effective AI collaboration lies in specialised agents trained on specialised data. Take Rovo by Atlassian as an example. This AI assistant has access to your entire work context - Confluence knowledge base containing strategy papers and PRDs, Jira projects showing current team work and progress, company goals, and quarterly planning artifacts. When you ask it a question about your work, its responses are grounded in your specific organisational context.
This represents a fundamental shift in how AI can support our work. Rather than generic responses based on internet-wide training data, these specialized AI teammates can provide insights that are directly relevant to your organization's specific challenges, priorities, and ways of working. Companies that understand and act on this distinction - investing in AI systems that are deeply integrated with their knowledge bases and workflows - will gain significant competitive advantages in how their teams operate and make decisions.
While specialised AI tools are becoming more prevalent, the reality is that most product teams aren't using AI in their daily work.
The key to bridging this gap is developing an "AI-first mindset" - approaching each task or challenge by first considering how AI could help tackle it more effectively.
It's not about using AI for everything, but rather building the judgment to know when and how AI can genuinely enhance your work.
Create sustainable AI habits
By moving employees beyond basic AI usage to strategic collaboration, companies can capitalize on AI’s full potential and drive significant ROI. But shifting your AI mindset is not just about time savings – it’s much bigger than that. AI collaboration can give companies an additional competitive edge by promoting continuous improvement, elevating work quality, and fueling meaningful innovation.
The key to meaningful AI adoption lies in creating sustainable habits.
Throughout the day, you might use it as a sounding board, exploring different approaches to user problems or refining your communication to stakeholders.
Here are some practical ways and examples to build these habits:
During product discovery
Use AI to analyze interview transcripts for patterns you might have missed, generate fresh hypotheses for testing, and identify emerging themes in user feedback.
For example, tools like Sauce AI can collect feedback from multiple sources and analyse them to inform decision making, identifying themes and common challenges raised across community groups.
In product strategy
Leverage AI to explore different market scenarios, analyze competitive landscapes from new angles, and generate alternative strategic narratives.
Tools like Perplexity Pro allow for incredibly fast research across a variety of sources. It doesn't do the research for you but gives an entry point from which you can decide where to go next.
AI teammates often work best at the start or end of a workflow – at the beginning of an ideation process to help spark ideas, or at the end of writing to help tighten up content. Always verify sources from any research you outsource to AI agents.
During product development
As a Product Manager, there are many daily workflows you can delegate to AI:
- Writing product requirements
Using specialised agents can save hours that you can reinvest in high-value activities.
- Drafting release notes
AI can generate release notes from Jira issues by identifying common themes across issues, presenting a summary and list of released issues in a problem-solution format.
The key is to use AI as a partner during these processes, not as a replacement. This helps ensure that AI enhances creativity rather than diminishing it, addressing the job satisfaction concerns seen in the MIT study.
Anticipate where your AI teammates will evolve next
The research from MIT gives us valuable clues about where our AI partnerships might head. For product managers specifically, here's what to watch for:
The rise of specialized AI
We're likely to see increasing commoditization of underlying models, with more processing happening client-side. To remain competitive, AI capabilities will become more specialized and move "up the stack," offering fine-tuned and personalized capabilities to both developers and end-users. We have seen this already happen with tools like Cursor, which demonstrate the transformative potential of specialized AI by embedding context-aware intelligence directly into developers' workflows, turning general AI capabilities into purpose-built assistants that understand both code and intent.
The winners won't necessarily be those with the most powerful models, but those who best integrate AI into existing workflows and user experiences.
Mixture of Experts (MoE) models
A Mixture of Experts model is like a company with specialist departments.
Instead of sending every question through the entire system like traditional LLMs, it has a "router" that quickly decides which "expert" parts of the model should handle the task. This means only the relevant parts activate for each specific job - like using physics experts for science questions and language experts for writing tasks.
The benefit? Much greater efficiency at lower costs, allowing you to build more specialized, efficient products that can handle diverse tasks without breaking the bank.
The human element in Product Management
Perhaps most importantly, as AI becomes ubiquitous, people will increasingly crave authenticity, quality and meaning. We'll likely see:
- A renaissance in high-touch customer service and craft
- Greater value placed on world-class storytelling and emotional engagement
- A resurgence of small businesses offering unique, personalized experiences
- New opportunities for leveraging uniquely human capabilities
The ultimate winners in this new era will be those who understand that while the science of business is about scale, the art of business often lies in the things that don't scale – the human touches that AI cannot replicate.
What this means for Product Managers
We've discussed how PMs can work with AI to make their jobs easier, but there's also the question of how PMs can better build AI into their product offerings. Success will likely require:
- Understanding how to orchestrate multiple AI capabilities rather than relying on single models
- Focusing on interface design and user experience as key differentiators
- Building personalization into core product experiences
- Maintaining the human elements that give products meaning and emotional resonance
- Creating workflows that enhance rather than replace human capabilities
We're all still figuring things out. And of course the enabling technology is a moving target as it continues to improve. There's been real progress, but we're still just at the beginning of this journey.
The most important thing is that we now have a better sense of the right questions to ask, and how to go about determining the answers.
The journey ahead isn't about becoming obsolete or becoming an AI expert. It's about learning to be a better product manager in a world where AI is just another tool in our toolkit. And while that journey might sometimes feel uncomfortable, it's one worth taking.
