Overriding AI results and manual ranking

Overriding AI results and

manual ranking

The AI is a recommendation, not a verdict. Here is how to override scores, suppress recommendations, and pin candidates the way you want them.

You can always override, and the system expects you to

Every AI output in Talinty is overridable. Scores, recommendations, the 3-column brief, the order of the candidate list. None

of it is locked. If the AI says Hold and you say Advance, the candidate moves. If the AI puts a candidate at rank 47 and you

put them at rank 1, the team sees them at rank 1.

This is the right default for two reasons. The AI is working from the criteria you set and the data the candidate provided; both

can be incomplete. And hiring decisions are consequential, which means a human owns them. The product is designed to

make the human in the loop the easy path, not the friction-filled one.

Four kinds of override

Move a candidate against the recommendation. Click the stage selector on the candidate card and choose the stage you

want. The AI's recommendation stays visible on the card so the rest of the team can see you overrode it, but the candidate is

in the stage you put them in. No confirmation prompt, no warning, no friction.

Pin a candidate to the top of a list. Open the candidate card, click the pin icon. Pinned candidates appear at the top of the list

regardless of their AI score. Useful for hand-picked candidates from sourcing, referrals you want the team to see first, or

anyone whose score does not reflect the reason you want them considered.

Suppress the recommendation entirely. On the candidate profile, under the AI Hiring Recommendation block, click "Suppress

recommendation". The recommendation disappears for everyone on the team. Use this when the recommendation is

misleading enough that you would rather the team make the call cold than be primed by it.

Annotate the score. Next to either score, you can add a private note explaining why you disagree. The note appears on the

team's view of the card. The score itself does not change; the annotation makes your reasoning visible. This is usually a

better choice than suppressing, because it teaches the team how you read the AI rather than hiding the disagreement.

[Illustration: Four small UI patterns shown as cards stacked vertically. Each card shows a different override action: stage

move (arrow between stages), pin (pin icon active), suppress (recommendation block crossed out), annotate (small comment

chip next to the score). Signal White background, Forest typography, Talinty Green for the active controls.]

What the AI does with your overrides

It does not learn from them. Your overrides do not feed back into a shared model that affects other customers, and they do

not affect the AI's behavior on the next candidate.

This is a deliberate choice. It would be technically possible to learn from overrides, and it would be slightly easier in the short

run. It would also create a system where past hiring decisions silently shape future hiring decisions, which is exactly the

failure mode that hiring AI is criticized for. The Talinty AI is bounded by the Core Criteria you set, not by patterns it inferred

from your team's history.

If you want the AI to behave differently, change the Core Criteria. The Criteria are the lever. Overrides are the safety valve.

When to override versus when to retune

Two heuristics that hold up:

Override when the candidate is unusual. A referral with a non-standard background, a candidate who interviewed last year

and is reapplying, someone whose profile reads thin but whose work samples are exceptional. The AI does not have the

context you have. Override and move on.

Retune when the pattern repeats. If you are overriding the same kind of candidate three times in a week, the Core Criteria are

wrong. Go back to the criteria, fix them, let the pool re-rank. Override is fast for individual cases; tuning is the right answer

when the AI is consistently missing the same thing.

If you find yourself overriding more than twenty percent of the candidates the AI ranks highly, the criteria need a serious

rewrite, not more overrides.

Overrides and the audit trail

Every override is logged. Who did it, what they overrode, when, and any note they attached. The log is available in the

candidate profile's activity feed and at the workspace level for Admins.

This matters in two situations. The first is internal: the team can see who made what call when the role is reviewed. The

second is regulatory: in jurisdictions where AI-influenced hiring decisions require explainability (EU AI Act being the live

example), the override log is part of how Talinty stays compliant. The recommendation is explainable; the human decision is

auditable; the gap between them is documented.