Turning black-box AI rankings into transparent, human-centered hiring decisions.
An AI-native applicant tracking system that helps recruiters evaluate candidates without giving up their judgment. This solution proposal brings match quality, confidence, and explainability into one coherent experience.
The system could generate a shortlist in seconds, but adoption remained below 30%. Recruiters could not understand why one candidate ranked above another, hiring managers felt out of control, and compliance teams lacked defensible explanations. Speed alone was not solving the problem.
Hiring manager persona · Design research artifact
I reframed the experience around human–AI collaboration: make the reasoning visible, separate candidate fit from confidence in the evidence, and give recruiters meaningful control over the final shortlist.
Separate Match Score from Confidence Score so a strong fit is never mistaken for strong evidence. Surface missing information and explain why each candidate is ranked.
Allow recruiters to review, correct, and freeze a candidate’s position before a hiring-manager handoff. Explain the consequences of every consequential action.
Route low-confidence candidates into an improvement workflow rather than hiding uncertainty behind a single score. Preserve a traceable decision history.
A ranked list is not enough when the person making the decision cannot tell whether the evidence is reliable. The central UX task is to make a recommendation understandable, challengeable, and ready for a human handoff.
Recruiters, hiring managers, and compliance reviewers
A UX reading of the supplied artifacts. The sequence below explains the design logic, not a verified chronology of research. Intended benefits and proposed tests are not measured outcomes.
The persona artifact distinguishes recruiter speed, manager confidence, and compliance accountability. This keeps the brief from treating every stakeholder as a single generic user.
The workflow and journey maps connect candidate intake, evaluation, missing evidence, and handoff. Their purpose is to expose the decisions and exceptions between screens—not just describe the happy path.
The workspace brings match, confidence, alerts, and shortlist controls together. The freeze dialog then makes the consequence of stopping automatic updates explicit.
Next, ask recruiters to explain a ranking, identify weak evidence, and predict what freezing does. These are proposed validation tasks, not completed research findings.
A candidate can be a strong match while the available evidence is incomplete. Separate signals help people avoid equating a high ranking with certainty.
Two scores require explanation. Labels and evidence must make the distinction understandable without asking people to learn the model.
The freeze confirmation places the consequence beside the action: future recalculation will not change the frozen candidate. This supports informed commitment.
Confirmation adds friction. Reserve it for consequential changes rather than every routine action.
Missing information belongs in an improvement path, not hidden behind a confident-looking score. This gives the recruiter a next step.
Flags can become noise. Prioritize actionable uncertainty and avoid treating every missing field as equally urgent.
Connect the evaluation engine to the decisions people actually need to make.
The dashboard separates match, confidence, and risk. Ranked candidates sit alongside actionable flags and workflow progress, making it easier to review evidence before approving a shortlist.
Once a recruiter is ready for a manager review, a candidate can be frozen in the shortlist. A confirmation explains that future re-ranking and score updates will no longer affect that candidate, making the consequence explicit before the action.
These are the baseline and success criteria documented in the proposal, not claimed post-launch results.
Read the product requirementsExplore project presentations in Figma, without opening the working design file.
Recruiter workspaces, hiring workflows, and candidate journey maps.
Read the UX process & rationaleAI adoption depends on whether people feel empowered, not replaced. This proposal makes explainability, correction, and human judgment part of the core workflow—not an afterthought. The next step is to validate shortlist comprehension and confidence through usability testing, trust-calibration surveys, and audit simulations.
IndusInd BankA considered mobile experience that makes money management feel a little more human.