FlairX.ai
B2B
AI
SaaS
Startup
Conversational UI
0 to 1
Web
Reimagining Questionnaire Builder with an AI Copilot
Context
FlairX is an Interview-as-a-Service platform that helps companies scale technical hiring by outsourcing candidate interviews to vetted human experts. It offers end-to-end capabilities including job post creation, interviewer-candidate matching, and feedback workflows. While human-led interviews remain the USP, manual operations were holding back scale.
To address this, we set out to develop AI-enhanced workflows for interview template generation, AI-led interviews, and AI-generated feedback, beginning with the Questionnaire Builder.

My Role
As a UI/UX Designer, I contributed in:
UX Research
User Flows
Ideation
Wireframing
Mockups
Prototyping
Stakeholder Reviews
Team
2 UX Designer, 1 Product Manager, 3 Developers, 1 CEO
Tools
Figma
Miro
Canva
Notion
Google Docs
Timeline
2 weeks from getting the requirements till delivery
Skip to Iterations & Prototype
Overview
Problems & Opportunities
Manual Bottleneck
Admins were using ChatGPT externally to generate interview questions based on the job description, then manually entering them into FlairX.
What if we could
Leverage AI to generate editable question sets and skill categories from job descriptions?
Job Description Quality
Most job descriptions lacked structured details like skill priorities, time allocation, or clear must-have skills.
What if we could
Assign weightage and time allocation dynamically?
Inefficiencies
The manual process caused inconsistencies, misaligned interview focus, and poor candidate evaluations.
What if we could
Let users prompt AI to improve or customize questions?
Scalability
With growth, maintaining quality across hundreds of job roles and interviewers became unsustainable.
What if we could
Reduce turnaround time from 2.5 days to < 1/2 hour?
Research
Observations & Stakeholder Insights
Takeaways from discussions with internal teams

Original manual flow & why it wasn't working

Despite having a growing library of “master” and “company-specific” templates, stakeholders consistently avoided reusing them. Templates were perceived as generic, outdated, or misaligned with current role demands. They often started fresh or heavily modified existing ones, defeating the purpose of reuse.
Clients expected custom-crafted, high-difficulty questions that reflected their domain and hiring bar. Repetition, generic prompts, and low-complexity questions diminished trust and led to dissatisfaction.
There was no clear tagging, hierarchy, or smart recommendation system. Admins and interviewers struggled to find suitable question sets and often fell back on manual curation.
The system lacked intelligent assistance to help teams craft high-quality, unique questions or guide difficulty levels, resulting in time-consuming manual effort and quality inconsistencies.
Competitive Analysis
Platform
Strengths
Weaknesses
Notes

Video-based screening with AI scoring and facial analysis
Concerns around transparency, bias, and candidate trust
Over-reliance on visual/emotion cues

Human-led technical interviews with structured feedback systems
Doesn’t scale well for async or AI-first approaches
Heavily human-dependent

Real-time interview recording with live AI notes and recruiter insights
Lacks customizable templates or prompt-based question generation
Good foundation for AI note-taking

Limited question flexibility, rigid flow
Strong on rubric adherence, weak on customization

Automated proctoring, AI-based assessments, and plagiarism detection
UI not intuitive, limited feedback insights for interviewers
Great for coding but weak on collaboration

Structured interviews, async support, ATS integration |
Lacks AI-powered question generation and deep customization
Not designed for AI-enhanced flows

Real-time AI-generated interview notes and summaries; elegant UX
Lacks customization and template-driven workflows for different roles
Strong on passive AI capture, weak on AI-driven prep
- No platform provided prompt-driven question building with smart refinement tools.
- Most lacked human-in-the-loop design for feedback control, transparency, and trust.
Metaview and BrightHire are excellent at capturing interviews, but lack depth in AI-driven question creation.
- Our opportunity lied in building a modular, AI-assisted flow that balances automation with editorial control, ensuring questions are aligned to job descriptions and easy to personalize.
Desk Research
What the Law Says
AI must be disclosed to candidates upfront (e.g., NYC law).
Bias audits are mandatory for AI hiring tools.
Candidate consent is required if AI is used for evaluation.
Design Move
We used labels, tooltips, and edit controls to make AI involvement transparent and keep interviewers in charge.
What Ethics Recommend
AI should support, not replace human decisions.
All outputs must be editable, explainable, and rated for confidence.
Transparency should be built into the interface.
Design Move
We added “Why this suggestion?” links and visible confidence notes for all AI-generated content.
What Others Are Doing
Platforms like Metaview and BrightHire use AI to assist, not take over.
Human-AI collaboration is the new standard.
Who are
the users?
Navigating
Complexity
Novelty
This was the team’s first deep AI-led UX initiative, which made it difficult to predict what was feasible and how fast.
Timeline
Time was a constraint, we aimed to deliver the MVP in just one month, requiring design, engineering, and product to work in parallel.
Feasibility
There was extensive back-and-forth between design and engineering to validate what could realistically be built and how.
Collaboration
We held long cross-functional sessions to align on what mattered most, how to phase features, and how to make trade-offs.
Scope
Requirements were dynamic, evolving with technical feedback, legal considerations, and founder inputs.
Compliance
AI ethics and legality added another layer of complexity, we had to continuously research evolving AI laws and design for compliance.
Design &
Iteration
It was a complex flow with so many moving parts, so we began with mapping
the user flow

Things had to move quickly, so we created quick wireframes and iterated upon it
within a day


Four days & four rounds of ideation, designing and feedback led to the creation
of the AI copilot




Final Screens





A Glimpse of the Final Prototype

My
Learnings
Back to Top
SaaS
Janhvi









Get in touch!
Open to chat & collaborate.
Conclusion
This project demonstrated the power of human-centered design in addressing real-world challenges at the IMPACT pantry. By combining a physical categorization system with a streamlined digital platform, we significantly improved inventory visibility and task coordination—reducing item search time by 70% and increasing volunteer autonomy. Usability testing confirmed strong engagement across all age groups, with a SUS score of 71.7.
Looking ahead, we envision refining the system with multilingual support, offline capabilities, and real-world deployment to ensure broader accessibility and long-term adaptability in diverse pantry settin
AI
Startup
0 -> 1

