HiringThe reading assist
A second pass that explains itself. Your criteria, your weightings, your scale, plus a score against each one with the reasoning attached. The decision stays yours, and stays on the record as yours.
You write the rubric. There is no built-in idea of a good candidate.
Criteria in your words, a weighting on each, your scale, and your guidance on what a strong answer actually looks like for the work you do.
- Your criteria, in your words
- Your weighting on each one, summing to the whole
- Your scale, and your guidance on what a good answer looks like
- Your note on what the summary should cover
- One rubric for the organization: the same bar for every role
It shows its work. A number on its own would be worse than nothing.
Every criterion gets a score and the reasoning behind it, the overall is those scores weighted, and the run ends with suggested next steps rather than a verdict.
- A score against every criterion, with the reasoning for it
- A weighted overall out of a hundred
- A summary that reads like a colleague’s note
- Suggested next steps, not a decision
- The model that produced it, and when
Still comparable in September. Because the run keeps its own rubric.
Improving your criteria later never invalidates the assessments you’ve already made, because each one kept its own copy of the framework.
- The rubric is copied onto the run, not referenced
- So is the posting it was judged against
- Refining your criteria never rewrites an old assessment
- Runs from March and runs from September are still comparable
What it doesn’t do. Worth saying out loud.
It does not decide
Moving somebody forward, or not, is an operator action and it is recorded as theirs. The assist produces a score, its reasoning and suggested next steps: nothing in the product acts on them.
It can fail
A run has three states, and one of them is failed. When it fails the row says so and keeps the error, rather than quietly showing no score and letting you assume there was nothing to say.
One rubric, for now
There is a single framework per organization today: the same bar applied to every posting. Per-posting rubrics and a template library are known next steps, not something we are claiming.
One application, fully read
The criteria you wrote, the score against each, and the reasoning that got there: the whole assessment on one screen, with the decision still waiting for a person.
Read it, then act on it. In the pipeline, as a person.
Yours. You write the criteria, the weightings, the scale and the guidance on what a strong answer looks like. There is no built-in model of a good candidate: if your rubric says a trade ticket outweighs years served, that is what it weighs.
Not today. There is one framework per organization, applied to every posting. Per-posting overrides and a library of rubrics are the known next step; we would rather say so than imply the product already does it.
Nothing. Each run stores its own copy of the rubric and of the posting it judged against, so an assessment from March still says what it said in March. That is the point of the snapshot: a rubric you keep refining would otherwise quietly rewrite your own history.
Only if you want it to. Set it to run as applications arrive, per posting or as the default for new ones, or leave it off and run it on demand. When it runs automatically nobody is recorded as having asked for it, because nobody did.
It reads the application against the posting: the answers somebody gave to your questions. It is not searching the web for them, and it is not scoring them against other applicants.