The first five minutes matter.
The brief asked me to validate two feature ideas for Naholo. The interview pointed somewhere else. I followed the evidence and designed a focused onboarding layer for an AI issue tracker.
Use the web app, find meaningful pain points, validate them with a real user and prioritise. Two opportunities were suggested: team token stats and an internal social layer.
Audited the product, turned both suggestions into testable hypotheses, ran a behaviour-first interview and separated evidence from assumptions.
Neither suggestion held up. Onboarding came up unprompted, so I reframed the problem, designed a guided first-use journey and wrote the test plan to prove it.
Welcome to your workspace
Get oriented, then create something useful.Context & brief
Naholo is an issue tracker for AI-assisted software development. Its users are developers who want to move fast with AI while keeping issues, decisions and work traceable, and avoiding “vibe coding” and low-quality output.
In scope
UX issues in the web app. The CLI and AI skills were explicitly out of scope.
How it was judged
Not broad coverage: how I collect evidence, develop hypotheses and prioritise.
Suggested ideas
Team token stats and an internal social network. I treated both as hypotheses to test, not conclusions.
Research approach
I used the brief as a starting point, then deliberately tested whether its suggested opportunities matched a real user's behaviour.
- ExploreUse a fresh project and inspect an existing one.
- FrameTurn friction into user-centred hypotheses.
- InterviewAsk neutral questions and follow real behaviour.
- SynthesiseSeparate evidence from assumptions.
- PrioritiseChoose the direction with the strongest signal.
Token discoverability
Users may struggle to find where tokens live, understand their balance and cost, or know when tokens are being used.
Internal social layer
Users may need a lightweight way to communicate inside a project, like announcements or context around completed issues.
One conversation, four behaviour-first probes
I never asked “is this feature missing?”. Asking about real workflow gave the participant room to reject a hypothesis, which is exactly what good research should allow.
“Where would you expect to find token information?”
“Tell me how you work with other people on an issue.”
“Which part of your first use was least clear?”
“If one part could be improved, what would you choose?”
What I learned
Participant
Active Naholo user who mostly works independently and rarely needs project-wide communication.
Tokens
No prior token experience and didn't know where tokens were, so token questions couldn't be validated through lived behaviour.
Communication
Uses other tools when needed: sometimes ChatGPT first to organise thoughts, then Slack.
It can be great to have a better onboarding process for new users.
No token experience. A discovery signal, but not enough to call token UX a major pain point.
Mostly solo work. In-product communication had little value in this workflow.
Raised independently, unlike the two pre-supplied ideas.
Communication is spread across tools; ChatGPT is sometimes used to organise thoughts before Slack.
I needed time to understand how Naholo works. For a solo developer starting from scratch, that is a first-use risk.
Prioritisation
I weighted evidence quality, user impact and confidence rather than choosing the most interesting idea.
Suggested by the participant and consistent with my own first-use friction. Strongest direction to validate further.
Possible signal, but the participant wasn't a token user. Needs the right segment first.
Could matter for team workflows later. This interview showed no strong need.
Evidence changes the roadmap
- EvidenceOnboarding came up without prompting.
- InferenceNew users may lack a stable mental model.
- DecisionTeach the first useful task, not every menu.
- SolutionProgressive checklist and contextual tips.
- TestMeasure unaided first-task completion.
Design direction
Not a generic product tour. A contextual orientation system that reduces “what is this?” moments, then gets out of the way.
- LandA short overview says what the tool is for.
- OrientShow how issues and projects connect.
- ActCreate an issue with one useful example.
- DocumentIntroduce logs when they become relevant.
- ReturnGuidance stays available, never mandatory.
Clear hierarchy
One prominent first action instead of a full product tour.
Contextual help
Explain issues, logs and tokens only when they become relevant.
Visible exit
Skip, dismiss, revisit. A finished guide collapses into a quiet help entry.
Create an issue
Welcome back, Maya
How I'd validate it
Proof comes after the prototype. These are the tasks and metrics that could confirm or reject the direction.
Create a first issue
Without guidance: time taken and points of confusion.
Explain issue vs log
Can the participant describe both in their own words?
Find help again
Can they recover the guidance after dismissing it?
What I can and can't claim
One participant recommended onboarding unprompted. The token and social hypotheses weren't validated by that participant. I also experienced first-use uncertainty myself.
That onboarding is the highest-impact issue across users, that these screens solve it, or that any activation metric improved. All new UI is concept work.
Better research changes the direction.
The useful result wasn't agreement with the brief. It was a better next question, a narrower design response and a plan to test it.




