Designing Trust & Efficiency into AI Hiring
How I shaped Talliant's AI Interview Bot, Proctoring Warnings, Evaluation System, and Automated Scheduling to feel seamless for HR while protecting candidate integrity.
Senior Product Designer & Strategist
End-to-end ownership of AI feature integration, HR/Admin/ Candidate information architecture, and interaction patterns.
Ongoing platform evolution
Core AI capabilities designed and validated across multiple sprints.
AI Interview Bot (Mira) with real-time proctoring, timestamped warnings sidesheet (with duration filters), and Integrity integration
Automated transcription, call insights, and evidence-backed candidate evaluation forms
Refined Interview Setup flow including Slot Booking (duration selection, multi-panelist, multi-slot, Import Job reuse)
Drag-and-drop Workflow customization (Manual vs AI steps)
Candidate Journey summaries with ATS integration
Elimination of manual Admin → Panelist → Candidate slot coordination friction
A virtual AI hiring partner, not another database.
Talliant is an Agentic AI-powered Talent Engagement Platform that acts as a virtual AI hiring partner. Unlike traditional ATS tools that mainly store data, Talliant actively manages the recruitment lifecycle — screening resumes with intelligent FitScore ranking, engaging candidates through automated outreach and scheduling, conducting structured AI interviews via its agent Mira, running proctored assessments with integrity warnings, generating transcriptions and call insights, creating evaluation forms, and delivering real-time hiring intelligence.
It automates up to 90% of the recruitment funnel while keeping human judgment at the center of final decisions, helping teams hire faster, more fairly, and with greater transparency.
Recruiters were drowning in repetitive work — and still worried about integrity.
From research and the existing product flows, three core problems stood out.
Interview Integrity Was Invisible or Overwhelming
Video interviews happened, but there was no reliable way to surface cheating signals (looking away, script reading, impersonation) without creating extra review work or false alarms for HR. Warnings needed to be actionable, not just noise.
Evaluations Were Inconsistent and Time-Consuming
After every interview or call, someone had to transcribe, summarize insights, and build evaluation forms. The output varied wildly between recruiters. AI could generate this, but it had to feel transparent and editable — otherwise HR would never trust it.
Scheduling Was Still a Manual Nightmare
The flow Admin → Panelist → Candidate for choosing interview slots involved endless emails, calendar checks, and reschedules. Even with a Calendar view, there was no streamlined way inside Interview Setup to define duration, add multiple panelists, select precise time slots, or reuse setups from similar jobs.
Existing solutions — traditional ATS plus bolted-on AI tools — either hid the AI too much or exposed raw scores without context. Neither built the trust required for high-stakes hiring decisions.
Four rules, fixed before a single wireframe.
These principles directly influenced how I approached the proctoring warnings, evaluation forms, and slot selection flow.
Transparency Over Magic
Every AI output — warning, score, insight, evaluation — must show its reasoning and be editable by a human.
Cognitive Load Reduction
Surface only what HR needs right now. Warnings and insights should accelerate decisions, not create new dashboards to monitor.
Human-in-the-Loop by Default
AI proposes, HR disposes. The interface should make override actions feel natural and fast.
Seamless Journey Continuity
From job posting → application → AI interview → evaluation → scheduling, the system should feel like one continuous story, not disconnected tools.
Understanding the current state.
I started by mapping the provided role-based UI flows — HR Dashboard, Job Portal, Candidate Journey, Create New Job, Calendar. The existing architecture was already strong on modularity, but the AI outputs were fragmented across different panels. HR had to jump between “AI Call Insights”, “Interview Transcription”, and “Candidate Journey” without a clear narrative thread.
The manual slot selection pain was visible in the gap between the Calendar widget and the actual coordination logic — it was still very human-driven.
Redefined flows
I redefined the key userflows to create continuity. These were validated against the existing HR and Super Admin screens to ensure they felt native to the current information architecture.
Job Creation → AI Setup
HR gets AI assistance on role requirements and suggested interview topics/difficulty. This feeds directly into question generation and later evaluation criteria.
Interview → Evaluation
Mira conducts the interview → real-time transcription + proctoring runs in background → structured Evaluation Form auto-generates with Integrity Flags section → HR reviews, edits, and decides in one screen.
Shortlist → Schedule (Refined Slot Booking)
Inside the new tabbed Interview Setup (Job Details → Choose Workflow → Setup → Slot Booking), HR selects interview duration, adds or imports panelists, defines date/time ranges, and picks multiple available slots. The system surfaces a clean calendar and generates confirmation links. Panelists receive availability requests with success modals.
Choose Workflow, live
The tab those three flows share. HR drags Manual or AI steps onto a single canvas — ATS screening, Mira, phone calling, an else-branch rejection mail — so the job never feels like disconnected tools.
Key design decisions.
AI Proctoring Warnings that build — not break — trust
I designed a dedicated Warnings sidesheet, reachable from the Interview Transcription or Evaluation view, that shows a clean, filterable list: warning types clearly labeled with occurrence counts, timestamped entries with exact start/end times and duration, and quick actions to investigate or clear individual warnings.
The duration-dependent filters let HR filter warnings by interview length, so short interviews don't get drowned in noise. The sidesheet feeds directly into the structured Evaluation Form's Integrity section — scannable and actionable instead of overwhelming, while still giving HR full visibility and control.
↓ Try the duration filter
Evaluation forms that feel intelligent, not automated
The generated form now includes:
Competency scores with supporting quotes pulled directly from the transcript.
Separate “AI Insights” and “Integrity Flags” blocks.
Clear, one-click editing capability so HR can easily adjust or add their own judgment.
HR told us in testing that this single change made them actually want to use the AI output instead of ignoring it.
Solving the manual slot selection problem
I designed a dedicated Slot Booking tab inside the tabbed Interview Setup flow:
Duration selection with clear chips (15 min, 30 min, 1 hr, 2 hr, 3 hr) so the system knows exactly how long each interview block should be.
Panelist management — add multiple panelists by name + email, or Import Job to instantly reuse the entire setup from a previous similar posting.
Precise slot selection — date range pickers + time range, then multi-select available time slots in a clean list or calendar grid.
Location toggle (Online / Offline) with address field when needed.
Panelist-side confirmation flow with calendar availability picker and success modal.
This completely removed the old Admin → Panelist → Candidate email tennis. HR now configures everything in one place; the system handles invites and confirmations, and the Calendar view and Upcoming Interviews widget stay in sync automatically.
Role-based information architecture
Looking at the Super Admin, Admin, HR, and Candidate screens, I reinforced a clear separation. This prevents the common mistake of showing every AI signal to every user.
Gets actionable panels — AI Transcription, Insights, Journey — focused on speed and decision quality.
Gets oversight and configuration tools without being overwhelmed by candidate-level AI outputs.
Stays clean and focused on status and next steps.
What the platform delivers.
of the recruitment funnel automated while preserving human judgment
higher precision in screening vs. traditional human or ATS-driven methods
time saved by eliminating paperwork and scheduling chaos
AI interviews successfully conducted (L1 assessments)
diverse resumes processed with context-driven screening
of recruiter time on repetitive L1 interviews can be reclaimed
Plus a significant reduction in candidate drop-off — addressing the industry benchmark of 46% of candidates losing interest within two weeks due to slow communication.
What we chose, and what it cost.
Over-monitoring anxiety
Real-time proctoring alerts are powerful but can create over-monitoring anxiety. We kept them optional and defaulted to post-interview review.
Automation vs. senior roles
Full automation of slot selection is ideal for speed, but some senior roles still need manual panelist input. We made smart suggestions the default with an easy “Request manual coordination” fallback.
Speed vs. recruiter intuition
AI-generated evaluations are fast, but over-reliance could reduce recruiter intuition. The strong edit/override affordances and visible reasoning help mitigate this.
Designing AI for high-stakes domains needs a different mindset.
Trust is earned through editability and explanation
Not just accuracy. The proctoring warnings only became valuable once HR could easily understand, challenge, and contextualize them.
Information architecture matters more than the model
The same warning or insight can feel empowering or overwhelming depending on where and how it appears in the flow.
Automation should remove pain, not judgment
The slot selection and evaluation systems work best when they accelerate the human decision rather than trying to replace it.
The best AI interface often feels boring
In regulated or sensitive spaces — calm, clear, and predictable beats clever every time.
I'm proud of how we turned potentially scary AI capabilities — real-time eye tracking, automated evaluations — into features that feel protective and professional.
Want to discuss how these patterns could apply to your hiring tools or AI products?
I'm always happy to walk through the flows, trade-offs, or next iterations.
All information synthesized from Talliant's official product (talliant.ai), provided UI flows, and direct design work on the platform.