If you're leading a product or engineering team using AI coding tools, you've probably noticed something paradoxical: while individual developers are more productive, team coordination is increasingly slowing you down ... the friction increasing. Why?
The viscous friction problem for your feature team
Think about fluid dynamics for a moment. When you move a spoon slowly through honey, there's resistance but it's manageable. But try to move quickly through that same honey, and the resistance becomes overwhelming— this is viscous friction (which is directly, inversely a function of the velocity of movement).
The same thing is happening to your product development process. AI coding is increasing your development velocity, but your developers and whole feature team is moving through the same environment (the honey) and so the friction has increased. Your coordination processes and tools are still built for the old, slower world.
Let's consider three sources of friction and context loss:
Handoffs between people: Traditional feature development, even in "agile" environments, involves many handoff points between Product Management, UX, and Engineering. Each handoff happens between different people and different systems, creating opportunities for context loss and delays. For example:
- Product defines big picture one pager feature in a Google Doc → UX researches flows → PM aligns on features in Aha → UX creates wireframes in Figma → PM reviews design via a meeting and Slack → PM writes detailed requirements in Confluence → Engineering provides feedback on requirements and designs in Confluence → UX finalizes specs in Figma → Engineering plans architecture in separate Confluence → Engineering defines task and estimates in JIRA → Engineering builds prototype and code lives in Github → UX reviews internally → Engineering builds final version → QA testing → PM prepares launch docs in Google Docs → Deploy
Tool Fragmentation: Your context is scattered across many different tools. One pagers in Google Docs, features in Aha, requirements in Confluence, designs in Figma, tickets in Linear or Jira, code in GitHub, and developer notes in random text files that disappear when the feature ships. The more tools the more friction and opportunity for context loss.
Manual Syncs: Someone, usually a project manager or scrum master or engineering leaders, spends their days trying to keep everyone aligned on what's being built, what's done, and what's next. A human developer might spend time tracking down the product manager to ask clarifying questions.
This is the viscous honey that you are moving through and it is what is resulting in friction and context loss.
How AI makes viscous friction worse and context loss more of an issue
Communication Becomes the Bottleneck
Great product development is highly collaborative and iterative. But iteration requires constant communication about trade-offs, priorities, and changes. When development cycles shrink from weeks to days, this communication overhead becomes crushing.
Teams find themselves spending more time in status meetings than building features. Project managers become air traffic controllers, frantically trying to coordinate work that's happening faster than they can track.
AI won't do the work a human will to get the right context
AI-coding and prototyping makes it all the more important for the full context to be available and up-to-date for the developer. A human may be willing and able to spend the time finding out what is accurate or not and asking around for updates and context. An AI will not.
Using Claude Code to MCP to JIRA and Confluence is powerful but only a partial solution as you are leaving too much interpretation up to the AI.
Documentation Debt Explodes
AI coding encourages experimentation. Developers can quickly try four different approaches, keep the one that works best, and throw away the other three. But where do you track those experiments? How do you capture what you learned? How do you prevent the next person from repeating the same failed approaches?
Traditional project management tools weren't built for this kind of rapid iteration and experimentation.
A New Way of Working
The solution isn't to slow down development. We need to collaborate differently.
Shared, accurate context is king
This means:
- Developers and AI that understand full project context, not just individual tasks
- Requirements, plans, and todos available where the actual work happens
- AI tools and human team members accessing the same context simultaneously
- Updates flowing between specs and implementation
- Collective learning from AI experiments and iterations
Lightweight, continuous processes are needed
This means moving away from formal handoffs, ticket management, and heavyweight documentation toward:
- More frequent, lightweight check-ins with working prototypes
- Living documents that evolve with the code rather than getting outdated
- Iterative prototypes, experiments, and finished code
What This Looks Like in Practice
For my own team at Stravu, instead of separate PRDs, design specs, and technical documentation, we work from unified, collaborative planning documents for each feature. (Of course we use Stravu on Stravu to do so!)
These documents are leveraged simultaneously by Claude (our AI coding assistant), our developers, and me as PM. When someone discovers that a requirement needs to change, or when we learn something new during development, it gets updated in one place and flows to everyone who needs it. We are integrating this with some of the standard project management tools.
The result? We're building better features faster, with less coordination overhead and fewer miscommunications.
It is a lot more fun to swim in water than to try to swim in honey.
Learning together
AI coding isn't going to slow down. If anything, it's going to get faster and more capable. The teams that thrive will be those that redesign their collaboration practices around this new reality.
We are all figuring this out as it happens. What coordination challenges are you facing with AI coding? What changes has your team made to adapt to AI-powered development?
Stravu provides a way for feature teams to work together on unified, collaborative plans as they iterate and develop with AI.
If you are interested in exploring this with us, sign up for Stravu's Beta today!