Limits of Talinty's AI: what it does and doesn't do

Limits of Talinty's AI: what it

does and doesn't do

The honest list. Read this once and you will know exactly where to trust the AI and where to lean on your own judgment.

What the AI does

Talinty's AI does six specific things. Everything else in the product is conventional software (databases, workflows, integrations) that

does not involve a model.

Parses resumes, including Arabic right to left, into structured fields.

Scores candidates against the Core Criteria you set: Resume Match and AI Skill Match.

Generates structured vetting briefs (the 3-column view) explaining the scores.

Recommends a next step on each candidate: Advance, Hold, or Future Roles.

Drafts text when you ask: job descriptions, rejection emails, interview questions, summaries.

Answers your questions in context through the Talinty Assistant, reading the screen you are on.

That is the complete list. If something in the product behaves intelligently and is not on this list, it is rules-based automation, not the

AI.

What the AI does not do

This is the list users actually want when they search for AI limits. It is direct on purpose.

It does not reject candidates automatically. Every rejection requires a recruiter to click a button. Even auto-reject thresholds on

skills assessments produce a flagged candidate, not a deleted one.

It does not send messages without your approval. The AI drafts. You send.

It does not score candidates on protected attributes. Age, gender, nationality, religion, ethnicity, marital status, and pregnancy

status are excluded from the model's input. Names are evaluated only for parsing, not scoring.

It does not learn from your individual hiring decisions in ways that affect other customers. Your overrides, rejections, and

advances do not feed back into a shared model.

It does not predict job performance. It evaluates fit against criteria you set. Performance is what happens after the hire.

It does not interview candidates. Async video interviews are recorded by candidates and transcribed by the AI. The questions

and the judgment of the answers are yours.

It does not replace compliance review. WPS, Nitaqat, EEO, and equivalent requirements still need a human signing off. The AI

surfaces relevant fields; it does not certify them.

It does not work without Core Criteria. A job with no criteria produces no Resume Match and no Skill Match. The AI is not a

generic ranker; it ranks against what you told it to look for.

[Illustration: Two adjacent columns titled "Does" and "Doesn't". Six checkmark items in the Does column in Talinty Green. Eight xmark items in the Doesn't column in a muted Sage Gray. Forest typography. Signal White background. The columns deliberately

equal in width to signal that the limits are as important as the capabilities.]

Why this list exists

Most help docs in this category are vague because vagueness is safer for the vendor. The honest list is more useful for the user. If

you know what the AI cannot do, you know what you are responsible for, which makes the AI easier to trust on the things it can.

The Talinty AI is built to be explainable, auditable, and bounded. Those are the words we use because they are accurate, not

because they sound good. The same list above is what a regulator would expect to see, which is the second reason it exists.