Agentic AI Product Model: Rethinking the Product Operating Model Towards AI and Speed

We need to rethink the Product Operating models in these new times of speed, shortcuts and automation.

TL:DR Summary

Core shift: Old models rely on subjective meetings with opinions as output and gut feeling. The new model uses AI agents to continuously align strategy, user needs, priorities, and development.

The model includes examples of Agentic AI's working seamlessly in a loop to support faster discovery and development:

Synthetic User Agents (AI personas based on real user research) Strategy AI Agents (guide direction based on product vision, principles and business) Prioritization Agents (auto-manage backlogs and roadmaps always aligned) Development Agents (support building, always with the Synthetic User in the loop)

Goal: Fewer rituals, more real-time collaboration. Like when GPS replaced maps, AI will guide product teams with live data.

Result: Faster, smarter product work — with humans still in the loop.

Foreword:

I wrote this article based on my experience in product development, discovery, and ownership, working with Scrum and Agile since 2011 - and before that, building software for global brands using Kanban and looser processes. I'm a certified Product Owner and have been doing some form of discovery for 26 years.

My motivation came from a deep interest in people, product operating models, and the rise of Agentic AI. It's clear that new ways of working no longer fit old governance structures. So I wrote this piece.

When I finished, I felt the need to give the model a name - something that reflects adaptability. "Liquid" came to mind. Then I remembered Morten Elvang posted an article a while back on Liquid Organizations. His take on adaptability and change was surprisingly and coincidentally aligned with my thinking. Even though i have never read his article. So Morten we apparently share a common mindset. And I'm glad that my take on Product operating models in a change of AI somehow can fit into a "liquid" model. Even though our visions for future product organisation are not completely aligned.

Because of AI - how we build products is fundamentally changing.

The ones who will win / survive the next era of speed, AI and "winners take most" are NOT the ones with the biggest budgets, best brands or the best data. It's the ones that have the best adaptability / change readiness.

you are most likely stuck in:

day-to-day operation in old or current "operating models" in your "ways-of-working" or process of doing things

These barriers are keeping you from rethinking how you should work to get ahead of the curve in this "winner takes most" era.

Disclaimer: First we need to be open-minded and imagine that AI WILL eventually create great value both for your end users/clients/customers but also for your organisation. I am not going to, try to sell AI to you. Some are past that point. And if you still need convincing then someone probably will do the job for me the next coming weeks.

If you do know your way around the AI potential but are still skeptical then you should consider that it will get faster, more precise, less prone to hallucination, more sustainable and easier to use over the next coming months.

So let's talk about Product Operating Models going into this new AI / Speed era.

Today's product operating models are full of rituals that are meant for alignment. Planning, backlog grooming, priority meetings with leadership teams, retrospectives, helicopter meetings or trio meetings, check-ins, morning stand-up meetings.

Alignment that could be semi-automated because in all its essence it's about decision-making and next steps. Hopefully, mostly backed by some kind of data (as seen in the bullet list below).

It is all about removing subjective meetings with opinions as output and driving automatic data-driven alignment.

And now we can consume endless amounts of internal and external data instantly in our decision-making. So we need to go from biased human gut feelings to data-augmented decision-making. Data like:

goals and objectives direction and strategy ways of working company history and principles auto-updating roadmaps research on user interviews real-time always updated competition analysis / blue ocean strategy AI persona agents (synthetic users) what's in the backlog of all feature teams

Boil all that together into an "AI product agent" that can eliminate vague priority decision meetings and give clear recommendations to the trio — or soon-to-be duo maybe. Where the designers and developers of small feature teams use all this compiled data to be the strategic "human-in-the-loop". The AI agent provides the clear picture of what to build next. Get help generating new ideas for features based on all this data.

It's like when we used maps in the car — maybe you are old enough to remember all the copiloting and discussions on how to get to the destination (the product vision). Then came GPS — full automation. It completely removed all friction and solely focused the driver on driving and enjoying the ride.

So with AI and real-time internal data in the loop, the next move will always stay updated, clear and relevant for the current state of the product. Just like GPS.

The Liquid Product Model

Discovery

It's not about letting the AI decide what to build. The product designers/researchers, together with the engineers will still be the human deciding factor. Real user interviews will still (if possible) drive the opportunity landscape. But from the time of the interviews until the wireframes for the solutions are drafted, that time will shrink tremendously in the coming months and years.

The way we drive human research will stay somewhat the same. User research will include more about human factors between man and machine in terms of trust and reliance. Designers will start to shape the AI interface as instructions rather than pixels and button interactions. In other words, prompt engineering is the new UI work. Not to build web/app interfaces or prototypes, but work on creating AI agents and their system prompts that define their behavior UX uses AI to analyze interviews (meeting calls and notes) Designers can use text-to-design to get off to a solid draft head start before editing the real design into it

Synthetic AI User Agents ( AI personas ) are deployed using all gathered data so the rest of the team can "talk to the research data".

AI User Agents that:

Are set up with the perspective of the real archetypes / personas Can simulate the pain points and is biased and affected by the symptoms of the real users. Are informed by empathy maps/mind maps, journeys and affinity maps. Can convey in a real way what the days of a user look like. Are built to answer to both product strategy and development as well as product designers. Can also be set up to answer to other "product AI agents" leading to the next chapter.

Product Management

Product management is the keeper of time and priority. They will still be the people person that drives timing.

But the delivery cycles will change from 1-2 week sprints to continuous everything : "Continuous Discovery" and "Continuous Delivery" with fewer rituals.

What will change is the sources of information. Right now opportunities land in the backlog or opportunity-solution trees from discovery, to be handled by product owners over the course of several alignment meetings, to get to development in JIRA. This work will end up being semi-automated enabling one product manager to do the work of several product owners, by governing the product agent pipeline of data from discovery through the curated product model to end up as a prioritized ticket in the auto-sorted backlog. This might sound very futuristic and visionary and to some impossible. Because as long as we have humans in the loop then there will be management/facilitation of some sort.

PMs use "AI product agents" trained on gathered discovery data, business strategy, the product vision, your business objectives and key results and much more. - to help prioritize what is the next move based on the the current roadmap.

More of the rituals going forward will be in line with the flat organisation structure we see rising at the moment, which means the rituals might be more about culture, trust-building and psychological safety in teams working even closer together - compared to now where they only touch base in morning stand-ups and other rituals. More time together on the field also means they need to trust each other and respect each other's boundaries. This happens through trust-building and teamwork.

This leaves us with great opportunities to rethink how we work faster while not burning out. And how we run discovery, both synthetic and real, with human factors at the forefront of what is important.

Development

We go from cross-department deliveries and JIRA handovers to overlapping ways of working more than ever before. Less time spend discussing what definition of ready and definition of done is and writing tickets. More time spend building products.

We're moving from handovers to collaboration. From specialists working in silos to cross-functional teams building together. The roles will stay the same. But with more AI products coming, semi-automating more and more tasks in production and products is important.

More and more products will move towards AI automation or semi-automation with "human-in-the-loop" and this kind of development surfaces a lot of new ways of working for developers and designers.

The synthetic user agents can be hooked up to your virtual coding agents that will know who to build for before suggesting any changes Prototypes are built by designers and developers — in the same tool, almost at the same time. Both shaping details by hand but also prompting to shortcut standard cases, icons and images AI assisted-coding is used for faster boilerplate kickoff for developers Developers will be wrapping data services for MCP (defining the interfaces between the data and the AI), while also using AI workflows and working more on node-based interfaces

Roles and ratios

👤 1. User Research Role Ratio: 1 per 10 developers

Drives human-centered research, informs AI user agents, and validates synthetic feedback loops. Still essential for grounding teams in real human needs.

🧠 2. Technical Research Role (System Architect) Ratio: 1 per 20 developers

Ensures product agents, models, and pipelines are fed with structured, clean, and connected data. Defines interfaces and integrations between internal systems and AI agents.

📍3. Product Manager Ratio: 1 per 20 developers

Oversees the AI backlog and roadmap agents. Ensures they're aligned with real-world delivery needs and intervenes where strategy, timing, or constraints shift.

🧭 4. Strategy Lead Ratio: 1 can cover most teams in a midsize organisation

Ensures all product agents operate within the company's broader vision, strategy, and ethical guardrails. Shapes direction across squads without micromanaging details.

🎨 5. Product Designer Ratio: 1 per 1-4 developers

The designer brings aesthetics, interaction logic, and human empathy. Their "backpack" includes design principles, UI systems, and storytelling. Works with the offset of prompting a good base design system and layouts - Then work the design in place in Figma or a like. The files then gets shared with the code environment, and based on the structure of the Figma file, AI helps rebuild the layout in code.

💻 6. Developers

Builds the product's core. Developers ensure quality, security, and scalability. Their "backpack" includes architecture patterns, performance awareness, and coding craft. They collaborate with designers and AI agents to bring ideas to life quickly and cleanly. They also govern the build pipeline and the code workspaces. Also assisted by AI workflows.

We don't need more rituals. We need to find a way to stay in momentum, stay fluid like flowing water — or a car with GPS. You don't stop to discuss the next path ahead, with trust in the AI product agents (governed by a strategic "human-in-the-loop"). It keeps using data to stay informed all the time, not just after meetings. It's basically about deciphering why we hold meetings and then automating those priority landscapes within several layers of static and real-time data. Most of it we already have at hand from the list above.