> hice matching shows criteria, evidence and doubts before the score: you know why a candidate fits, what's missing and what to ask in the interview. Source: https://hice.ai/en/chat-ai/candidate-matching # Choose the right candidate, and know why they're the right one. When you're looking for someone for a project, hice puts what you're asking for on one side of the scale and the evidence it finds on the candidate on the other: career, interviews, CV. It shows you what fits, what's missing and what still needs checking. The score comes last, and you can redo the math yourself. Matching is in every plan, at no extra cost for AI. ## Criteria first, then evidence. The score comes last. - 1Your request goes on the first panJava, London, available from January, fluent German. Until there's evidence, the scale tips that way. - 2The career comes inLuke Foster has worked in Java for six years, the last two as technical lead. What counts is when and for how long, not just that the word Java appears in the CV. - 3The interview comes inFrom the notes of the person who met him: London works for him, and he can start on January 7. - 4One doubt remains, and you can see itOnly the CV says his German is fluent: nobody has heard him speak it yet. The scale almost balances, and German stays on the list to check at the interview. ## First, it reads the request The first thing hice looks at is what you're asking for. The job title leads, the description refines it. Each requirement gets one of three weights, and it doesn't change weight for convenience. If a requirement is ambiguous, like “senior” without saying how many years, hice doesn't decide at random: it keeps it as a stated threshold and lets the evidence speak. If you want to clarify it, just write it in the chat. “We're looking for a senior Java developer for our ERP migration, in London, from January. German is a plus.” ### Must-have Required. If it's missing, it's missing: no high score makes up for it. ### Preference Welcome. It weighs on the result, but on its own it rules nobody out. ### Inference It isn't written down, but the context makes it clear: “migration” means knowing both the old system and the new one. It stays marked as an inference and never becomes a must-have. ## Then, it reads the profile What counts isn't how many right words are in the CV. It's what they prove. Here's what it looks at, on an example CV. ## When the sources disagree Seven sources speak about a candidate, and they don't all weigh the same. The order is set in advance, not case by case: the longer the bar, the more the source weighs. - 1The structured interview assessment (the scorecard) - 2Notes from the person who ran the interview - 3The career you can verify: dates, roles, companies - 4Certifications - 5The text of the CV - 6What the candidate states in screening forms - 7The AI's inferences, which never decide on their own ### A difference between sources isn't averaged: it's declared If the interviewer writes “basic German” and the CV says “fluent”, the criterion follows the higher-ranked, more recent source. And the report shows you the conflict, with both statements. ## The report: every result with its evidence Not just a percentage. Every criterion has a status, every status cites the evidence behind it, and the score is a sum you can redo by hand. - “2018–now, Java developer, Beta Inc.” · careerhigh quality - “One month's notice” · screening formlow quality - What day can you start, notice period included? - Tell me about the last client meeting you held in German. ### Contradicted means conflict Two sources say opposite things. Nobody picks one in silence: you see who says what. ### “To verify” is an honest result Better a stated doubt than an invented certainty. Doubts aren't rounded off to make the number work. ### Two candidates side by side It explains what you gain and what you lose by choosing one or the other. It doesn't change the ranking: it helps you understand it. ## You decide the priorities Put each criterion in a priority tier and mark the mandatory ones. Two criteria in the same tier count the same. Change the priorities, and the shortlist is recalculated. You can look at the same request from different angles, ask for five more profiles, dig deeper into a candidate and dismiss a suggestion you don't need: once dismissed, it doesn't come back. ## When the answer is no, it tells you If no candidate covers a criterion, the report says so and shows you what would widen the shortlist. If the analysis isn't finished, because some profiles haven't been assessed yet, it says so instead of passing it off as complete. A weak shortlist stated clearly is worth more than enthusiasm with no basis. None of the 24 profiles has experience in the banking sector. If you remove it from the priorities, 5 more profiles come in. ## On your next search, look at the evidence before the score. Matching is in every plan, at no extra cost for AI. If your candidates are in spreadsheets or another ATS (the software for managing hiring) today, on the Business plan you bring them in with bulk migration. - Free for up to 10 people - No credit card required - Data migration with Business ## Read alongside matching - Candidates and recruitingFind the right person by searching skills, not folders. - CV in chatFrom file to profile in seconds, always with your OK. - Staffing and recruitingFrom candidate to consultant on assignment, in one tool. - A reliable assistantIt sees only what you can see, and never makes things up.