Explainable mentor matching based on skills and goals.
MentoraHQ scores every possible pair against criteria you set, shows you exactly why each one scored what it did, and waits for you to confirm it. The scoring is deterministic — the same inputs produce the same score every time. AI writes the explanation for each pair. It does not choose your pairs, and nothing here learns from your outcomes.
58
matching dimensions in the library to build your criteria from
6
weighted scoring dimensions, with weights you set that sum to 100%
100%
of pairs confirmed by a human administrator before anyone is notified
0
models trained on your participants' data
How the mentoring matching algorithm actually works.
Four steps, all of them inspectable. If you cannot explain a match to the person who received it, the match is not finished.
Step one
You choose the criteria
Pick from a library of 58 matching dimensions — industry, function, job level, experience, location, time zone, language, management responsibility and more. A typical participant questionnaire uses 12–18 of them, because a 100-question intake form is how response rates collapse.
Step two
Candidates are retrieved and scored
Every eligible mentor is scored against every mentee across six weighted dimensions: questionnaire fit, engagement, availability, goal alignment, experience and diversity. You set the weights; they sum to 100%, so a change to one is visibly a trade against another.
Step three
AI writes the rationale
This is the only place a language model is involved. It reads the score breakdown that the rule engine produced and writes the plain-English reason this pair ranks where it does — including the dimensions where the pair scored badly. It never changes a score.
Step four
An administrator confirms
Recommendations are a ranked shortlist, not a decision. You confirm, swap or reject any suggested pair, and nobody is notified until you do. Bulk assignment proposes a whole cohort at once, and you review it the same way.
What “AI” means here, precisely.
Most mentor matching platforms will not tell you which half of this they do. So here is the line, drawn in public.
What the AI does
- Writes the explanation for why a pair scored as it did, in sentences a participant could read
- Summarises a mentor or mentee profile so an administrator reviewing 200 candidates does not read 200 forms
- Flags where a pair looks strong on paper but weak on a dimension you said mattered
What the AI does not do
- Decide the pairing. The ranking comes from the deterministic rule engine, not a model
- Learn from your outcomes. There is no training loop on your participants' data
- Change over time without you. The same inputs give the same scores next quarter as today
This is a deliberate position, not a gap we are hiding. Outcome-trained matching needs years of completed relationships before it beats explicit criteria, and an unexplainable match is one an administrator cannot defend to the person who received it. Rules-based mentor matching that you can read beats a black box that you cannot — particularly where mentoring touches promotion, sponsorship or protected characteristics.
Mentor matching software for the way your programme actually runs.
Skills-Based Mentor Matching
Match on the skill a mentee is trying to build against the skill a mentor actually has, using a taxonomy that understands that a Cardiology mentee can be served by a general Medicine mentor but not the reverse.
Mentor Compatibility Assessment
Questionnaire responses from both sides are scored against each other rather than against an ideal, so compatibility reflects two real people and not a persona.
Automated Mentor Matching in Bulk
Propose an entire cohort in one pass, with mentor capacity respected so your most willing mentor does not quietly end up with nine mentees.
Availability-Aware Pairing
A pair that cannot find an hour in common is not a good match however well it scores on paper. Time zone and working hours are scored, not an afterthought at booking time.
Mentor Matching Criteria You Version
Criteria and weights are configuration, not code. Simulate a change against your real participants before you publish it, and see which pairs would move.
Student and Alumni Matching
The same engine runs university alumni mentor matching: match by course, graduation year, industry and career goal. See university mentoring.
Mentor matching questions
- Is this an AI matching algorithm or a rules engine?
- The ranking is a rules engine: weighted scoring over criteria you configure, fully deterministic. The AI layer sits on top and explains the result. We describe it as AI-assisted rather than AI-powered matching because the distinction matters to anyone who has to justify a pairing.
- How do I automate mentor-mentee matching without losing control of it?
- Automation here means the scoring and shortlisting, not the decision. You set criteria once, the engine ranks every candidate pair in seconds rather than the day per cohort that manual matching takes, and you confirm the shortlist. Nothing is sent to a participant until an administrator confirms it.
- Can I override a recommendation?
- Always, and the override is recorded alongside the score it replaced. Administrators know things the criteria do not — a reporting line, a history between two people, a pending move. The engine ranks; it does not outrank you.
- Does the matching improve as we run more programmes?
- Not by itself, and we would rather say so. Improvement comes from you adjusting criteria and weights in light of what you have seen, with a simulation step that shows the effect before you publish. There is no silent model drift to explain to your legal team.
- What data does the matching use?
- Only what participants supply in their profile and questionnaire, plus engagement and availability recorded inside the platform. Sensitive dimensions are optional to disclose, and where they are used for matching they are scored without being exposed to other participants.
See the matching engine on your own criteria.
Bring the dimensions that matter to your programme and we will show you the scores, the explanations and the overrides on real data.