What the AI does in five steps

When a candidate's application lands in Talinty, the AI runs a fixed sequence before you ever see them on the candidate list. There is no model deciding whether to act; the same steps run on every candidate, in the same order.

1. Parse the resume. Talinty extracts structured data from the candidate's CV: employment history with dates, education, skills, certifications, languages, locations. Arabic CVs are parsed right to left with the same fidelity as English ones.

2. Score the resume against the job. The AI compares the parsed resume to the Core Criteria you set on the job. The output is a Resume Match score between 0 and 100. Higher means closer alignment, not better candidate.

3. Score skill alignment separately. A second score, the AI Skill Match percentage, looks past resume keywords and reasons about whether the candidate's actual experience maps to the skills the role needs. This is the score that catches candidates whose CV understates them and the candidates whose CV overstates them.

4. Build the structured vetting brief. Behind each candidate, Talinty assembles a 3-column view: Why They Match, Bonus Strengths, and Things to Consider. This is the AI showing its work.

5. Recommend a next step. The AI synthesizes the two scores and the vetting brief into one recommendation: Advance, Hold, or Future Roles. The recommendation is a starting point for your judgment, not a replacement for it.

All five steps are complete before you open the profile. Nothing waits on you to click anything.

What "signals" actually means

You will see the word signals across Talinty. It is not marketing. Each of the five steps above produces a separate signal, and the platform deliberately keeps them separate rather than collapsing them into one number.

• Resume Match is one signal: how well the CV reads against the job, as a document.

• AI Skill Match is a second signal: how well the underlying experience maps to the work.

• Verified assessments (when attached to a stage) are a third signal: actual demonstrated skill on a task.

• Async video and live interviews are a fourth signal: how the person communicates, reasons, and behaves under light pressure.

The AI reads all four when they exist. Multi-signal hiring is the working theory behind the product: any single signal is gameable, but the combination is not.

Where the AI's reasoning ends

The AI scores, organizes, recommends, and explains. It does not decide. Every action that moves a candidate forward, sends an email, or commits the company to anything still requires a click from you or a teammate.

This is by design. The AI Act in the EU and emerging hiring regulations elsewhere expect humans to be in the loop on consequential decisions. Talinty's AI is built to be helpful inside that constraint, not around it.

For the full list of what the AI specifically does not do, see the Limits of Talinty's AI article.