Technical CPO breaks down the AI agents her team uses in project delivery
Learn how a CPO uses AI agents to automate PM tasks, speed up delivery, and focus on outcomes, with real examples of workflows that cut delays and improve team performance.
AI Maintenance: Product People uses AI to handle maintenance tasks, allowing focus on achieving business outcomes.
Human Touch: Empathy, judgment, and strategy remain human strengths AI cannot replicate in product management.
Agentic Solutions: Deploying AI Agents has accelerated workflows, improving speed and stakeholder satisfaction.
Low-Maintenance Tools: Client-agnostic tools like Notion and Figma support a distributed, efficient operational model.
Future PM Role: AI will redefine project management, with focus shifting to strategic and system design skills.
Mirela Mus is the founder and CPO of Product People. She has an MBA and a background in computer science and more than a decade of experience in product management — half of which she has also spent in a leadership role.
We connected with her to learn more about delegating maintenance to AI so that humans can focus on getting real results. Here's what she had to say.
Delivering business outcomes
I'm Mirela Mus, the founder and CPO of Product People. We are the premier destination for product management and product growth services in Europe. Some of our clients include Amazon, Zalando, eBay, DeepL, and the World Health Organization.
Unlock for Free
Create a free account to finish this piece and join a community of forward-thinking leaders unlocking tools, playbooks, and insights for thriving in the age of AI.
Have an account? Log In
While we agree that execution is a great moat, we don't just oversee or manage projects; we deliver business outcomes like increasing revenue and decreasing costs.
Why some elements cannot be replicated by AI
We view the AI-first evolution through our core principle: "Low maintenance, high results". AI handles maintenance so we can focus entirely on results.
We aggressively adopt tools that accelerate baseline execution, so that we can spend less time on the unglamorous hands-on work, like:
- Qualitative data synthesis: We use AI to synthesize public market data, user reviews, and support tickets instantly. This supports our "GSD (Get Sh*t Done) Attitude." We don't waste time staring at blank pages.
- Initial artifact drafting: AI is excellent at creating the first draft of PRDs or release notes. This supports our current methods, ensuring junior PMs start with a structure that meets our high internal bar, thus reducing the manual review cycle.
- Rapid context gathering: Instead of long discovery phases, we use AI to accelerate our understanding of new industries and market norms. This allows us to onboard fast and start delivering value immediately.
We view the AI-first revolution through our core principle, ‘Low maintenance, high results.'
With work automated, I double down on human elements AI cannot replicate:
- Deep empathy (uncovering the "why now?"): AI can analyze data, but it cannot read the room. A human must look a stakeholder in the eye and understand the real pressure behind a request — whether it's a genuine market need or just a panic reaction to a competitor. So, spend more time influencing stakeholders and navigating the political context that often blocks delivery in complex organizations.
- Unbiased judgment ("mercenary mindset"): We use the time saved to apply what I call a "mercenary mindset," removing emotion from decision-making to ensure clear judgment. This allows us to cut through internal noise and solve for the client, prioritizing their actual business goals over internal popularity contests.
- Vision setting and strategic betting: As I often say, "Only relentless optimists are good fits for Product Management." AI can predict from past data, but it cannot rally a demoralized team toward a better future or make the counterintuitive bets required to transform a company.
- Navigating complex organizational politics: This is the hardest part of the job. Dealing with departments dragging their feet or managing stakeholders with conflicting agendas requires high emotional intelligence and "mercenary" focus.
- Decision speed (OODA Loop): We use efficiency gains to tighten our "OODA Loop" (Observe-Orient-Decide-Act), ensuring we pivot strategies faster than traditional in-house teams.
AI can predict from past data, but it cannot rally a demoralized team toward a better future or make the counterintuitive bets required to transform a company.
How agentic solutions increased speed and satisfaction
We recently used agentic solutions when we faced a classic bottleneck. The time between ideation and development was bloated. Our PMs spent massive amounts of energy manually cross-checking legal constraints and chasing stakeholders for sign-offs. This high-maintenance work added no strategic value.
We deployed a system of internal AI Agents to handle the unglamorous hands-on work:
- The compliance guardrail: We configured an in-house agent using context from the client's internal legal and compliance knowledge bases. Before sending a PRD for review, the agent pre-validated it, immediately flagging regulatory risks. This served as an instant bar-raising check, ensuring the client’s PMs did not waste cycles on non-compliant drafts.
- The stakeholder chaser (via Slack): We automated the approval workflow. Instead of a PM using their limited political capital to nag a director for a signature, the agent handled the chasing via Slack. It relentlessly but politely sought approvals, removing the PM's emotional drain.
- Automated progress updates: We implemented an agent to synthesize progress and development data and send proactive status reports.
The results:
- Speed: We significantly reduced the average time between ideation and development. The "dead time" of waiting for approvals collapsed, and the agent, acting as a devil’s advocate, validated PRDs faster.
- Satisfaction: Stakeholder satisfaction increased. They valued the proactiveness and timeliness of the automated reports, freeing our PMs to focus on delivering outcomes rather than administrative nagging.
Why agentic experimentation is so important in project management
And here are a few more agents we've experimented with. We use Claude Code, Make.com, Glean, NotebookLM, n8n, and/or Notion AI.
- Artifact standardizer: Trained on the “Best PRDs” to reformat docs and Jira tickets. This is tailored to the product type, area (internal, external), and stage (maintain, invest, innovate).
- Devil’s advocate: Pokes holes in the PRD documents using compliance regulations and enriches user knowledge base indexes, competitor information, and company context. Saves five hours per week per PM.
- Dev kickoff: Reads Jira tickets and kicks off dev processes by setting up the environment and repositories, and pulling up the relevant libraries. Saves three hours per week per dev.
- Incident triage: Ensures timely and accurate relative prioritization for each incoming support or bug ticket.
- Stakeholder updater: Maps stakeholders and sends weekly digests and/or reports across the Product Funnel. Saves three hours per week per PM in selection and communication time. Here's an example of what a report might look like to a CEO: “Projected ROI down to €3m (-15%). This is because Initiative 1 is delayed by 3 months in Northern Europe due to a dependency in Initiative 2 and miscommunication with software vendor. There have been two VP level escalations to resolve it. Resolution ETA is one week.”
- Single source of truth: Employees ask Notion AI. It searches not only Notion but Slack, Google Drive, Jira and other connected applications.
I've personally also experimented with OpenClaw on a new Mac Mini.
How a low-maintenance stack supports a distributed setup
Our stack is client-agnostic — we work with whatever the client uses. Internally, however, we run a rigorous, low-maintenance stack that supports our working model and distributed setup.
Tool: Notion
- Category: Internal knowledge base and playbooks
- Impact: The single source of truth for our proprietary delivery model. It houses our entire Ways of Working, Policies and Enablement Functions.
- Evolution over the past year: Doubled down. We now use it not just for storage, but as the active training ground to "onboard fast" and deploy our specific methods (like "Optimism as an Operating System") to new hires instantly.
Tool: Figma
- Category: Design and prototyping
- Impact: It allows us to validate ideas via low-fidelity prototypes rather than lengthy specs.
- Evolution over the past year: AI Integration. We’ve moved from manual wireframing to using AI-driven template generation to speed up the early "messy" stage of design.
Tool: Miro
- Category: Visual collaboration and strategy
- Impact: The canvas for "Managing Chaos". We use it heavily for Opportunity Solution Trees (OSTs) and aligning stakeholders who are "pulling in different directions".
- Evolution over the past year: Standardization. We moved from ad-hoc brainstorming to strict, pre-set templates for workshops. This ensures every PM facilitates with the same "bar-raising" quality, even without a senior leader present.
Tool: Slack
- Category: Communication and community
- Impact: The backbone of our "kinship" and support network. It allows our PMs to "openly discuss problems" and get unblocked.
- Evolution over the past year: Disciplined Async. We shifted heavily toward "async" workflows to protect deep work. We use it less for real-time chatter and more for structured updates, reducing noise from political contexts.
How asynchronous video supports team coaching
Asynchronous video, via Loom and Miro TalkTrack, is one of the most underrated tools we use, because it enables our "Teaching Hospital" model at scale.
In our organization, "four-eye reviews" are critical to quality, but scheduling them synchronously bottlenecks the process. Now, instead of booking a 30-minute meeting to review a PRD or strategy deck, our Senior Consultants record a 5-minute Loom critique. This allows our PMs to get coaching without breaking their flow.
Additionally, stakeholders who are dragging their feet often won't read a five-page document, but they will watch a three-minute video. It forces alignment in complex political contexts where we need to bring order to chaos without adding yet another meeting to their calendars.
It perfectly embodies our "low maintenance, high results" mindset. We cut the meeting time by 80% while increasing the fidelity of the communication.
Why PMs must avoid the "cargo cult" of Agile
As for project management methods, we're moving away from heavy processes toward what I call the "Common Sense" framework. We explicitly avoid the "cargo cult" of Agile, where process exists for its own sake.
What we moved away from:
- Rigid planning: We avoid long-term Gantt charts and the feature factory mindset where success is measured by hitting the date rather than hitting the metric.
- Process bloat: We don't implement complex frameworks just because they are trendy. Many companies fail because their processes cannot support the weight they are forced to bear.
What we shifted to:
- Lightweight strategy: We use Opportunity Solution Trees (OSTs) and OKRs, but only sensibly. The goal is to align on the outcome (e.g., 'figure out something that will bring $10M ARR') rather than managing a perfect backlog.
- Operational excellence: In practice, we use a proprietary review process that assesses PMs' ability to deliver outcomes. We judge success by whether the client would work with us again or if they hit a major milestone, not by their documentation velocity.
How AI redefines project delivery rituals
We also have a few rituals that we have redefined with AI:
- Defining scope: AI handles the unglamorous, hands-on work of drafting standard feature specs. This frees the PM to apply a "mercenary" mindset, focusing entirely on the 20% of scope that drives the business outcome instead of managing a bloated backlog.
- Validating work: We use AI to instantly synthesize results from disparate tools, allowing us to move through the Observe-Orient-Decide-Act cycle faster. Instead of waiting a week to analyze a test, we can pivot immediately. It helps us navigate "VUCA" (Volatility, Uncertainty, Complexity, Ambiguity) environments quickly.
- Aligning teams: AI drafts routine communication summaries. This enables the PM to focus on "leadership without authority". We manage the political context and align stakeholders who might be dragging their feet to ensure the narrative lands with the people who matter.
How AI will redefine project management roles
Within five years, the role of the Junior Project/Product Manager will be almost entirely automated or redefined. Autonomous AI agents will manage routine project tracking, dependency management, basic documentation, and simple data-to-insight synthesis. The entry point for a human product leader will shift: You won't break into the industry by writing user stories; you'll break in by demonstrating exceptional skill in strategic thinking, system design, and human-centric leadership.
The PM role will fundamentally split into AI-Augmented Strategic Leaders (focused on vision and people) and AI-Augmented System/Process Architects (focused on designing the ideal workflow for the AI agents).
Why delivery leaders must shift their focus
Delivery leaders need to understand that our job is shifting from the doer of the process to the designer of the system.
So, here's my advice: Stop optimizing your own time on tactical tasks and focus on designing the AI-augmented systems your team will operate within. You need to:
- Codify your excellence: Clearly document your organization's proprietary knowledge and best practices, as this is the essential training data for your AI agents.
- Focus on the "why": Use the time AI saves you to become a better human leader, focusing on motivation, clarity, and navigating the ethical implications of the products your team builds.
The goal isn't to replace your team; it's to use AI to enforce your organization's definition of excellence at scale.
The goal isn’t to replace your team; it’s to use AI to enforce your organization’s definition of excellence at scale.
Follow along
You can follow Mirela Mus on LinkedIn and check out her company, Product People.
More expert interviews to come on The Digital Project Manager!