How to use FlowCanvas: four use cases from real discovery work
Live interviews, usability tests, stakeholder workshops and AI-assisted synthesis. How to set up FlowCanvas for each, with the shortcuts that matter.
,7 min read
- FlowCanvas
- Discovery
- User research
- AI
Most research notes die in a document. You type fast during the session, promise yourself you will clean them up, and a week later you are scrolling through a wall of half-sentences trying to remember which complaint came from whom.
FlowCanvas is built around one rule: capture first, organize later. During a conversation it gets out of the way so you can keep up. Afterwards it helps you turn what you captured into a map you can reason about and share.
Below are four situations where that pays off. Each one comes with a setup, the shortcuts that matter and a tip from using it in practice. If you just want to try it, open FlowCanvas and press N.
The five building blocks
Every capture is one of five types. You will use all of them in the use cases below.
| Type | Key | Use it for |
|---|---|---|
| Step | N | Something the person does: “Exports the report”, “Calls support” |
| Pain point | P | Friction, frustration, a workaround |
| Opportunity | O | An idea for how it could be better |
| Question | Q | Something to ask, check or follow up |
| Actor | A | A person, team or system involved |
Each capture holds a title, an optional note, an original quote and #tags. Captures connect as you go: with a capture selected, a new one of the same type continues to the right, and a different type hangs below. That simple rule is what turns a stream of notes into a map without any dragging.
Everything stays in your browser. There is no account, and nothing you type is sent to a server.
Use case 1: Capturing a live customer interview
This is what FlowCanvas was made for. You are on a call, the customer walks you through how they handle invoices today, and you need to keep up without breaking eye contact for long.
Setup. Start a new interview from the FlowCanvas start screen and rename it in the top bar (“Invoice flow — Maria, finance lead”). Stay in Interview mode, which is the default. It hides everything you do not need while capturing.
During the conversation:
- Press N and type the first step: “Receives invoice by email”.
- Press Tab for the next step to the right, already connected: “Downloads PDF”, Tab, “Types amounts into ERP”.
- She sighs about retyping numbers. Press P: “Manual retyping, errors at month end”. It hangs below the current step, so you will know later exactly where in the flow it hurt.
- She says something you want to keep word for word. Press Shift+Enter to move from the title to the note and on to the quote field, and type it as said.
- Type
#month-endanywhere in the text. It becomes a tag when you leave the capture. - Something to ask later? Q: “Who approves invoices above 5k?”
Afterwards. Switch to Review mode. Run Auto layout to tidy the map, open the inspector to fix typos, and use the Timeline view to replay captures in the order they came up. Insights shows counts per type, the most frequent tags and the balance of pain points to opportunities. It is a quick way to see whether the session was mostly problems or already full of ideas.
Do not organize during the interview. If a capture is in the wrong place, leave it. Moving things around costs attention you need for listening, and Auto layout fixes most of it in a second.
Use case 2: Taking notes in a usability test
In a usability test you watch someone work through tasks. The structure is already given: task by task, step by step, with problems along the way.
Setup. Create one interview per participant (“Checkout test — P3”). Before the session, add an Actor (A) for the participant and a short note with their profile, so the map makes sense to someone who was not there.
During the session:
- One chain of Steps per task. Start each task with a step that names it (“Task 2: change delivery address”), then Tab through what the participant actually does.
- Every hesitation, wrong click or “hm, where is…” becomes a Pain point under the step where it happened.
- Use the quote field generously. “I thought this button would save it” is worth more in a readout than your paraphrase.
- Tag severity as you go, for example
#blocker,#major,#minor.
Afterwards. Click a tag such as #blocker to spotlight every capture with it and dim the rest. Across three or four participants this shows quickly which problems repeat. Export each session as a PNG for the readout deck, or as a .flowi file to keep the full data.
Keep task names in the first step of each chain. When you compare participants later, you can find “Task 2” on every map at a glance.
Use case 3: Turning a stakeholder workshop into a map
Workshops and stakeholder interviews produce a different kind of raw material: sticky notes, a shared doc, someone’s bullet list in a chat. You do not need to retype any of it.
Setup. Collect the notes as plain text, one point per line. Bullets and numbering are fine; FlowCanvas strips them.
Paste it in:
- Click on the empty canvas and press ⌘/Ctrl+V. Every line becomes its own capture, connected in a chain (up to 50 lines at once).
- If a capture is selected when you paste, the new captures take its type. Select a pain point first and paste the “what bugs us” list, and they all arrive as pain points.
- Now sort. Select a capture and press Shift+O to turn it into an opportunity, Shift+Q for a question, and so on.
- Tag by theme or by who raised the point:
#sales,#support,#legal.
Afterwards. Use Auto layout, then search with / to jump between mentions of the same topic. The tag filter shows who cares about what, which is often the most useful outcome of a stakeholder round. Where do sales and support see the same problem? Where do they contradict each other?
Paste early, sort later. It feels wrong to drop forty unsorted lines on a canvas, but sorting on the canvas is much faster than cleaning up the list first.
Use case 4: Using a map as context for AI
A good map is also good input for an AI model. FlowCanvas does not call any AI service itself, since your data stays in your browser, but it gives you a clean file to bring to the model you already use.
Why this works. A .flowi file is plain JSON. Every capture carries its type, title, note, original quote, tags and when it was captured, and the connections between captures are included too. That is far more structure than a pasted transcript, and the original quotes keep the model close to what was actually said.
How to do it:
- In the ⋯ menu or the command palette (⌘/Ctrl+K), choose Export .flowi.
- Attach the file to your AI assistant, or open it in a text editor and paste the contents.
- Tell the model what it is looking at, then ask for one specific thing.
A prompt that works well as a starting point:
The attached file is a FlowCanvas export of a customer interview. It is JSON.
Each node's data has a nodeType (step, pain, opportunity, question, actor),
a title, a description, an originalQuote and tags. Edges connect nodes in the order
of the customer's workflow.
1. Summarise the customer's workflow in five to eight steps.
2. List the three most important pain points. For each, quote the customer
using only the originalQuote fields and name the step it belongs to.
3. Suggest two opportunities that are not already on the map, and explain
which pain point each one addresses.
Do not invent quotes. If something is unclear, list it as an open question.
Other requests that work well with a map:
- Across interviews. Use Export all on the start screen to get every interview in one file, then ask which pain points come up in more than one interview.
- Draft a persona or job story, grounded only in the captured quotes.
- Write an opportunity brief for one tagged theme, for example everything tagged
#month-end. - Check your coverage. Ask which steps have no pain points or questions yet. These are gaps you may want to explore in the next interview.
Keep the human part human. Let the model sort, count and draft, then check every claim against the map. The quotes are there so you can verify, not so the model can sound convincing.
A note on privacy: once you upload a file to an AI service, that service’s terms apply. Remove names and sensitive details from the map before exporting if your research consent does not cover it.
Where to go from here
The shortcuts above cover most of what you will need. For the rest, press ? inside FlowCanvas for the full cheatsheet, or ⌘/Ctrl+K to search every command.
Have a use case we did not cover, or a workflow that works well for you? We would like to hear it: hello@andronik.ai.
Open FlowCanvas and start your first map.