Qualit
How can teams maintain human oversight without reviewing every AI action?
Qualit is a working product concept I designed and built to address this question, connecting system awareness and AI triage with human decisions and execution.
Self-initiated concept
・
0 ➔ 1 Product Design
Working Prototype
・
Design Premise
Give AI autonomy where appropriate
Make AI judgements understandable
Bring humans in where judgment matters
Software quality is managed across monitoring, CI/CD, code review and issue tracking with each solving an important part of the lifecycle, but leaving oversight and human decisions distributed across them.
I spoke with 2 QA leads, an Engineering Manager and a Program Manager who were beginning to explore AI-assisted development workflows, supported by secondary research.
65% say AI writes or accelerates at least 41% of their code.
Smart Bear | 2026
38% say reviewing AI-gen code takes more effort than reviewing the code by humans.
Sonar | 2026
50% of developers lose 10+ hours/week to non-coding workflow frictions which include finding information and context switching.
Atlassian | 2025
The Opportunity
This led me to explore the experience between these areas:
When should a human
intervene?
Qualit explores a connected quality workflow where AI can detect, investigate and act on software issues autonomously; while bringing humans in when risk, uncertainty or policy requires judgment.
Structured autonomy
AI operates within defined confidence, risk and policy boundaries.
Human oversight by exception
Routine work can proceed autonomously; while critical decisions remain human.
Accountable execution
AI actions remain explainable, traceable and verifiable through resolution.
SIGNAL → EVENT → DECISION → EXECUTION → VERIFICATION → RESOLVED
Interaction model: AI acts. Humans govern when needed.
Human override remains available throughout.