Platform·How it works

From an empty job post
to a decision you can defend

Here is exactly what happens inside TalentSage at every step, what the AI does on its own, what it only does when you ask, and what stays entirely in your hands.

Request a Demo

Create the job

Go live

Auto pre-screen

Deep analysis

Skill checks

Riya interview

Decision

How TalentSage AI platform Works

From Job creation to intelligent screening and screening to AI interviews and onboarding, TalentSage automates the complete hiring lifecycle while keeping humans in control.

Job Posted

Recruiters create and publish jobs across multiple hiring channels in seconds.

Candidate Applications

Candidates apply through careers pages, job boards, referrals, or direct sourcing.

AI Auto Screening & Smart Rejections

AI instantly analyzes resumes, skills, and job-fit criteria.

Auto-reject low-match candidates

Resume Parsing & Candidate Intelligence

Selected candidates are parsed into structured profiles with AI match scoring and insights.

Automated Interview Scheduling

Qualified candidates automatically receive interview invites based on availability and workflow rules.

Brain Test & Skill Verification

AI-powered assessments and tasks validate skills, knowledge, and problem-solving ability.

AI Interview with Riya

Riya conducts intelligent AI interviews, evaluates responses, and generates summaries.

AI Evaluation Results

Generate complete candidate reports with scores, insights, rankings, and risk indicators.

Human-in-the-Loop Approval

Recruiters and hiring managers review AI recommendations before final hiring decisions.

AI-Assisted. Human Approved.

Offer & Onboarding

Send offers, collect documents, and automate onboarding workflows for a smooth candidate experience.

70%

Faster Screening

Reduce time-to-hire

Recruiter Productivity

Save hours every week

85%

Automation Coverage

Across hiring workflow

92%

Match Accuracy

AI-powered precision

100%

Human Oversight

Ethical & transparent hiring

Create the job, two ways

Both routes live on the same screen and produce the same thing: a draft you review before anything goes live. You can switch between them while you work.

AI Chatbot

Default

You describe the role in plain language, for example a full time onsite Senior AI Engineer based in New York.
The assistant asks follow up questions and gives you inline pickers for industry and department, salary display, visa sponsorship, security clearance, and how many screening questions to generate.
The first time only, it asks for a short description of your company so the writing matches your actual business. That description is checked before it is accepted.
When the draft appears you can keep chatting to request changes, then approve it or reject it and start again.

Manual Form

Full control

You fill the fields yourself across job identity, work setup, compensation, application requirements, core requirements, and pre-screening.
The job title is checked as you type, so vague titles and two roles squeezed into one post get caught before anything is generated.
You choose how many screening questions to generate, from one to ten. Five is the default.
You review the finished draft in a preview, then approve it or reject it.

What the AI actually writes for you

A structured job description with sections such as overview, role overview, responsibilities, requirements, must have, technical skills, soft skills, nice to have, benefits, and compliance.
A screening question set, each question typed as free text, yes or no, rating, or numeric, with a difficulty and an optional knockout answer.
A quality score out of 100 with a breakdown, so you can see how the draft was rated before you publish it.
The application setup: which documents and links a candidate must provide, and which are optional.

Every section, section title, screening question, and requirement stays editable. Your edits save straight to the draft, and job sections remain editable after the job is live.

Publish it and start receiving applications

Approving a draft turns it into a live job and gives you a share link plus an embed code you can drop into your own site.

Candidates apply on your career page without creating an account.
A resume is required, and it must be a PDF under 2 MB.
They provide name, email, phone, location and notice period, plus optional salary expectations, an optional cover letter up to 1500 characters, any links you asked for, and answers to your screening questions.
They receive a tracking id and can check their own status later on a status page.
A job can be published, paused, closed, or archived. Paused and closed jobs stop showing on your career page straight away.

Every applicant is pre-screened automatically

This part runs on its own the moment someone applies. There is no button to press.

Your core requirements are checked first, covering education, work mode, work authorisation, experience, and salary. Anyone who fails is clearly flagged, and still visible in your list rather than quietly hidden.
The resume is then scored out of 100 across four areas: skills match, experience relevance, project relevance, and education fit, each carrying a weight you control.
You set the rejection threshold, which starts at 50. Anyone scoring below it is marked as rejected by AI automatically.
While it runs the status shows as pending, and if a score ever fails you can run it again.

Run a deeper analysis when you want one

The deep evaluation is deliberately manual. You open a candidate and run it yourself.

You get an overall score out of 100, a shortlist or reject recommendation, per area scores, risk flags, and a confidence level.
A written explanation covers strengths, gaps, risks, recommendations, and how ready the person looks for an interview.
Answers to your screening questions are judged one by one rather than as a block.
If the candidate shared a GitHub, LinkedIn, portfolio, Figma, or Behance link, you can run a Deep Dive on it. Those cards only appear for links they actually submitted.

Prove the skills with Brain Tests and Brain Tasks

You choose who to test. Both are generated from the job itself in a minute or two, and you can edit every question and the rubric before sending.

A Brain Test is a timed set of questions, mixing multiple choice, coding, and open ended, at a difficulty from entry level to expert, with a time limit between five and sixty minutes and a pass mark you set.
A Brain Task is a take home assignment such as a coding challenge, case study, design, written piece, data analysis, or presentation. It defaults to about four hours of work with a seven day deadline, and candidates submit a file, a link, a GitHub repository, or written text.
Candidates get a secure link, confirm their identity using the email and phone you already hold, and work in the browser while progress saves as they go.
There is one attempt only. Leaving the tab or running out of time submits the work, and copy and right click are switched off. There is no webcam on tests.
Everything is scored out of 100 against your rubric, with per skill scores and plain notes on strengths and gaps. Submitted code repositories are reviewed as well.
Skill Veracity then lines up what the candidate claimed against what they proved across tests, tasks, interviews, GitHub, and their resume, gives a veracity score out of 100, and flags any skill where the claim sits above the evidence with a severity and the size of the gap. Reports export to PDF.

Riya runs the first round interview

You pick who to interview, set the length, and choose whether to record and proctor the session. You can add a hint about what Riya should focus on.

Riya emails each candidate a secure link that stays valid for as many days as you choose, and you can resend it.
The interview runs in the browser as a genuine back and forth. Riya speaks out loud and the candidate answers by voice or by typing.
The camera and microphone stay on so the session can be recorded and checked for integrity. Anything unusual is flagged for a person to review.
Riya builds each next question from the previous answer, so it can follow a thread instead of reading a fixed script.
You receive a full transcript, an overall score out of five and per area scores with the reasoning behind each, strengths and risks, every question marked full, partial, or missed, a short summary, and a suggestion with a confidence level.
Interviews are tracked as invited, in progress, completed, no show, or live now, with the average length across them.

Your team makes the decision

The AI scores, ranks, explains, and suggests. It never advances anyone on its own.

Candidates sit on a board with columns for applied, under review, shortlisted, interviewing, offered, hired, future opportunities, on hold, decision pending, and separate rejection columns for before and after the interview.
Every move is made by a person, either by dragging a card or using the dropdown, which only offers the moves allowed from the current stage.
You can move or reject in bulk, and send the rejection email at the same time.
Every stage change is written to a full history, so you can always see who moved a candidate, and when.

Worth knowing before you start

The limits are as real as the features, so here they are in plain terms.

Resumes must be PDF and under 2 MB.

Riya interviews are conducted in English.

Tests allow a single attempt, and leaving the page submits the work.

Creating a job spends a job description credit, and analyses, assessments, and interview minutes each run on their own credit balance.

Offer letters and onboarding are not part of the product today. Offered and Hired exist as stages you move candidates into.

Common questions

No. The chatbot opens by default because it is faster, but the manual form gives you every field directly and you can switch between the two at any point while creating a job.

“Automation is only worth having when you can still see the reasoning. Every score here comes with its evidence, and every decision still belongs to a person.”

The TalentSage principle