We didn’t start with a feature.We started with a workflow nobody wanted to own.
Podcast teams weren’t struggling because they lacked project-management software.
They were struggling because production was fragmented across people, tools, messages, files, approvals, and repetitive operational work.
Northflow became our attempt to redesign the operating system around that work.
01
We thought the problem was organization.
The first version of the idea was straightforward: podcast teams juggle a lot, so give them one place to keep it.
Calendars, episodes, guests, assets, tasks. Put it in one system and the chaos goes away.
That premise was half right, which is the most expensive kind of right.
Early Northflow. A space per show on the left, active work as cards on the right, each one carrying its own progress, people, and date.
02
Customers showed us the real problem was operational drag.
In interviews, nobody described their problem as disorganization. They described the same handful of moments, over and over.
- Chasing a guest for a bio and a headshot, four times
- Re-explaining the same process to a new contractor
- Not knowing whether an episode was actually blocked or just quiet
- Doing the same twelve small tasks for every single episode
- Finding out something slipped only after it slipped
None of that is a tracking problem. Tracking it more carefully just tells you sooner that you are behind.
The work itself was the problem.
A better list of what you are behind on is still a list of what you are behind on.
03
We built workflows around production, not around tasks.
Instead of a generic board with cards on it, we shaped the product around how a production actually moves: spaces per show, a production overview that surfaces where things are stuck, guest intake that collects what you need without four follow-up messages, templates so nobody rebuilds the same episode from scratch.
The projects view was the first place this clicked. Every card answers a different question than a status column does: not what stage is this in, but what is holding it up, and who is holding it.
Inside one episode. The columns are the production stages, not generic To Do and Done, so an episode’s position in the board is the same thing as its position in the process. Guest sourcing empties before pre-production fills.
Production across every show in one place.
04
Then AI changed the question.
Instead of asking:
“How do we help teams track more work?”
we started asking:
“How much of the work can Northflow actually take off their plate?”
That reframing changed the roadmap more than any feature request did.
A tracker gets more valuable as you put more into it. That is a tax. If the product can do the repetitive parts of production instead, the value stops being proportional to how much you feed it.
So we moved toward specialized agents: narrow, scoped to one job in the production process, doing the parts that are repetitive and legible rather than the parts that need judgment. Drafting the episode description. Preparing the guest one-pager. Assembling the assets that always get assembled.
The test we hold them to is simple. Did a human have to redo it?
Interactive — assign the task to a person, or to an agent
Send guest prep kit
Goal
Ensure the guest is fully prepared for the recording.
- Share episode brief: topic, angle, and key questions
- Provide technical setup guide (mic, headphones, quiet room, lighting)
- Include a ‘day of recording’ checklist
- Ask guest to do a 2-minute test recording if possible
Details
- Reads the episode brief
- Drafts the four prep items
- Attaches the setup guide
Handed back. A drafted prep kit sitting on the task, waiting for a human to approve or edit — not sent to the guest on its own.
The assignee field is the whole argument. A person and an agent occupy the same slot, so delegating to an agent is the same gesture as delegating to a teammate — and the agent hands work back for approval rather than acting on its own.
Project management organizes the work.Northflow is being designed to help do the work.
05 What being a founder taught me
The parts you can’t learn from the outside.
Customer discovery is not a phase
It is the whole job. Every time we drifted away from talking to customers, we built something slightly wrong and found out slowly.
Positioning is a product decision
We changed what we said we were, and that changed what we built. The words came first more often than I expected.
Roadmap is subtraction
Almost every good decision was a decision not to build something. The requests were real. The capacity was not.
Staying close to the build
When you own the outcome, you stop handing off. I prototype in code because the fastest way to know whether an idea works is to use it.
AI is a scope question
The interesting question was never what the model can do. It was which slice of the work is safe to hand over.
Founders make the tradeoff visible
In a big company the tradeoff hides behind process. On a small team you say it out loud, to someone’s face, and then live with it.
Northflow is live and still changing.
Have a complicated problem?
Those are usually the interesting ones.