Most companies are "playing" with AI by using it to summarize meetings or rewrite emails. We decided to run a real-world test at Tappwater.
We took a product development project and assigned an AI agent as the primary Product Manager.
The agent's remit
- Create the PRD (Product Requirements Document).
- Source potential suppliers and manufacturing partners.
- Create RFPs for 5+ suppliers and select the suppliers for the pilot.
- Prepare for every supplier call and answer their follow-up questions.
- Provide weekly management updates and highlight project risks.
- Navigate compliance and regulatory issues.
- Propose product names and build the Go-To-Market plan.
- Respond to, and proactively follow up on, all communication with the suppliers.
The human role
The human PM didn't disappear. They attended the calls, recorded them, verified the reports, and copy-pasted the emails. They became the "Evidence-Based Filter."
The results? Mind-boggling.
The agent managed the process better than any human on our team could have. It didn't get tired, it didn't forget a requirement, and it didn't miss a supplier follow-up.
But here is the real ROI: we didn't fire the PM. We added an entire additional product pilot to our roadmap without hiring a single new resource.
We achieved "scale without headcount."
The "company memory" moat
The best part isn't the speed. It's the knowledge retention. Because the agent was the PM, it knows everything about the product. If a team member leaves, the knowledge doesn't walk out the door. It stays in the log. It stays in the system.
We used P30 for this, but the platform matters less than the architecture.
If you are a founder, stop trying to use AI to "assist" your team. Use it to augment your team's capacity. Stop hiring for the grunt work. Start hiring for the oversight.
The real product pilot is launching in November, but I'm convinced. This is how you actually build an AI-native company.