AI screening

Four hundred applicants. One ranked shortlist. Nobody read four hundred CVs.

Upload a CV and get a considered opinion back — scored, summarised, with the concerns spelled out. It recommends. You decide. Nothing is ever filtered out of your view.

Five things, all of them writing rather than deciding

A read on the person

A score out of 100 with its reasoning, a short professional summary, up to five strengths and five concerns, a seniority level based on real scope rather than years, and the two roles this person actually fits.

Expected salary, in their market

A realistic range in the currency and pay period of the market they work in, inferred from their location and history — with a comparable figure alongside it so you can price a cross-border role sensibly.

A read on the fit

A match score against one specific role's requirements, experience level and key skills, with a written recommendation — so a long list can be ordered by someone who can see why.

Interview questions from their CV

Categorised question sets where every question names a real company, project, technology, date or gap from that person's own history. Generic questions are explicitly disallowed.

Drafted outreach

A personalised email about the specific role, written for you to edit and send. It never sends by itself.

A job stability read

A low, medium or high assessment of movement between roles, with an explanation — because the explanation is usually what changes the decision, not the label.

And skills extracted automatically, so a CV becomes searchable structured data rather than an attachment nobody opens again.

Two hundred applications, one afternoon

01

Applications arrive. Each CV is parsed and analysed on arrival — score, summary, strengths, concerns, seniority, expected salary, skills.

02

You open the role and sort by match score. The reasoning sits next to each number, so you can see which ones the model misjudged — and it does misjudge some.

03

You read the concerns lists on the top thirty. In practice this is where most minds change.

04

For your shortlist, generate interview questions grounded in each person's actual history — the prep a good interviewer does and a busy one skips.

05

Draft outreach to the ones you want to move fast on. Edit, then send.

06

Everyone you didn't progress keeps their analysis on their record, so the work isn't repeated when they apply again or surface for another role.

Nobody was filtered out. All two hundred are still there, still searchable, still in your database.

What you need

Your own AI provider account. You choose the provider and model and supply your own key, so your candidate data is covered by an agreement you've already vetted and you can see the cost per request in your own account. Three providers are supported.

Or nothing at all. Don't configure a key and these features are simply absent — everything else in the product works exactly the same. Some teams can't send candidate data to a third party, and that's a supported way to use this.

Each plan includes a monthly allowance of analyses; beyond that they're charged per analysis.