The AI ATS that does matching for you

Upload CVs, write a brief, AI returns a ranked shortlist. The traditional manual interface is always available for those who prefer it.

Why classic ATSs aren't enough anymore

Traditional ATSs have become static CV databases: keyword search, dozens of filters, zero context. Recruiters spend hours doing what AI would do in seconds.

Keyword-only search: type «React» and miss «ReactJS» CVs
Duplicate CVs and outdated versions polluting the database
No automatic matching between brief and candidate
Manual one-by-one outreach
Client reports built by hand in PowerPoint

Candidate matching as a conversation

hice.ai reads every CV, understands it and connects it to briefs in real time. You ask for a profile, it proposes the top 10 with reasoning.

«React Senior, 5+ years, fintech, Milan» → 12 ranked matches with match reasoning
«Need an agile PM, fluent English, healthcare» → shortlist with 0-100 scoring
«Reconnect candidates who said no to Generali 6 months ago» → AI re-proposes them
«Send personalized outreach to the top 5 of this shortlist» → drafts ready to review

Try it in chat

Examples you can ask:

Static demo of hice's AI chat.

What you get

Multi-format CV parsing

PDF, Word, images, scans: AI extracts name, skills, experience, certifications.

Semantic matching

Understands synonyms, equivalent stacks, real seniority: not just keywords.

0-100 scoring per brief

Every candidate has a score with explanation: why it matches, where it doesn't.

AI deduplication

Recognizes the same candidate even if the CV is different, no duplicates.

Candidate memory

Interview history, rejection reasons, preferences: all recallable in chat.

Personalized outreach

Drafts tailored to each candidate, keeping the recruiter's tone.

Per-search pipeline

Configurable stages: screening, interview, offer, hired.

Shareable shortlists

Client links with interactive shortlists and inline feedback.

Automatic skill tagging

Up-to-date skill taxonomy, no more manual tagging.

AI client reporting

Search dashboard generated and updated automatically.

Multi-tenant for agencies

Per-client data isolation, granular permissions.

Traditional interface

Prefer classic ATS with lists and filters? Always available alongside AI chat.

What a matching engine actually compares

Automatic matching has a bad reputation because the first generation compared literal keywords: a CV that said "Kubernetes" scored, one that said "container orchestration on K8s" did not. Semantic comparison fixes that specific failure, and it is worth knowing that this is the part it fixes, and not everything else.

What it does not resolve on its own is the requirement quality. A vacancy written as a wish list of twenty must-haves produces a shortlist of nobody, or a shortlist ordered by who padded their CV best. Matching improves in direct proportion to how honestly the role separates what is required from what is desirable.

Semantic comparison of skills, not literal keywords
Separation between required, desirable and trainable
Availability and location weighed alongside skills
An explanation of why each candidate ranks where it does

Bias, explainability and what to document

Any automated candidate ranking has to be auditable, and that is a practical requirement before it is a legal one: if a recruiter cannot explain to a hiring manager why one profile is ahead of another, the shortlist will be ignored and everyone goes back to searching by hand.

The workable approach is to keep the model out of the decision and inside the ordering. It proposes a ranked list with the reason for each position; a person decides who advances. Recording who reviewed, what they changed and why creates the audit trail that matters when a decision is questioned months later.

Visible reason for every position in the ranking
Human decision recorded on every advance or reject
Fields excluded from scoring, documented as such
Periodic review of results by role and seniority

Keeping the candidate database from decaying

A matching engine is only as good as the pool it searches, and a pool decays fast: skills grow, availability changes, salary expectations move, and someone who said no in March may say yes in October. A database nobody refreshes returns confident answers about a world that no longer exists.

The cheapest maintenance is to attach a freshness date to the volatile fields and let the engine down-weight what is stale rather than hide it. A candidate last contacted eighteen months ago still belongs in the results, flagged as needing a check, and that flag saves the recruiter the call that starts with outdated information.

Freshness date on availability and expectations
Stale profiles down-weighted, not silently hidden
Re-contact prompts for strong profiles gone quiet
Outcome of every process written back to the profile

Recruiting agency, 8 recruiters, 60 active mandates

An agency with 8 recruiters on 60 searches used a classic ATS: 50% of time on CV search and screening. With hice.ai AI handles parsing, matching and first shortlist. Avg shortlist time: from 2 days to 2 hours. Mandate capacity: +60%.

hice.ai vs traditional ATSs

Feature	hice.ai	Classic ATS
CV search	AI semantic	Keywords
Automatic matching	Yes	No
Candidate scoring	0-100 explained	Manual
Deduplication	AI	Manual
Personalized outreach	AI per candidate	Generic templates
Reporting	Automatic	PowerPoint
Learning curve	One chat	Weeks of training