Methodology
Brief. Map. Shortlist. Land.
Four stages, every search, every time. The discipline shows up in what we will not do as much as in what we will.
Brief
Three structured conversations with the hiring committee. The first is wide — what does the role look like at the eighteen-month point, who would do it brilliantly, what would cause us to decline a candidate at second-round. The second is narrow — required vs nice-to-have, comp band sign-off by the budget owner in writing, hiring committee locked. The third is mechanical — process, scheduling, weekly cadence.
The output is a written brief signed by the budget owner. Until that signature, we do not approach a candidate.
Our internal AI tool — Atelier — is set up against the rubric in this stage. Atelier is not seen by candidates and is never the basis of an in-or-out decision. It is a screening aid for our consultants when bulk inbound CVs arrive on a posting.
Map
The named consultant on the search builds the long list of 30 to 60 named candidates. Of these, roughly a third are people we know already; the remainder are sourced through targeted outreach.
First-round qualifying calls run across weeks 2 to 6. Typical first-round call is 45 minutes — 15 minutes on the candidate, 15 on the role, 15 on fit and motivation. We move calibrated candidates to second-round inside the same week.
We hold a weekly 30-minute review with the hiring manager throughout this stage. We expect them to attend personally.
Shortlist
A shortlist of four to five named candidates is presented to the hiring committee with written assessment notes against the brief. The notes name what each candidate does and does not bring to the role; they are honest in both directions.
On-site loops run in weeks 6 to 8. We coordinate scheduling, debrief with both sides after each round, and surface concerns the moment they arise. A candidate who has gone quiet for 48 hours after an on-site loop is flagged to the hiring manager personally.
Our hard rule: no candidate who has not been spoken to by their named consultant in the last five working days is presented to a hiring committee.
Land
Reference work runs in parallel with offer-shape negotiation. We do not chase references through HR portals — we ask the candidate for the names and call ourselves. Five references is the standard, including one investor or board reference where the role is at director level or above.
Offer-shape negotiation is structured around the candidate's personal context — equity-event timing, tax residency, family stability, planned tenure — not just the headline base. We have written about this on our /insights page.
After the offer is accepted, we stay in touch with both sides through the first 90 days of the placement. We will tell the hiring manager when their new hire has stalled out and we will tell the new hire when their hiring manager is uneasy.
What we will not do
- · Run a search before the brief is signed by the budget owner.
- · Take more than three concurrent retained mandates per consultant.
- · Subcontract the candidate-facing work on a senior brief to a junior consultant.
- · Use a model or any other automated tool to reach an in-or-out decision on a candidate.
- · Approach a candidate's current employer for a reference before they have an offer in hand.
- · Run volume hiring or graduate hiring on either practice.
A note on the AI tooling we use internally
Caliber operates an internal CV-screening tool called Atelier. Atelier is used by our consultants as a triage aid on inbound bulk CV submissions — typically 60 to 200 CVs per posted role. The tool runs a structured, deterministic ranker over CV data extracted from PDF, scoring against the role rubric on five dimensions, and surfaces a ranked shortlist alongside flagged risks (visa fit, salary band, employment-history gaps, over- and under-qualification).
Atelier is bound by three rules. First, a personally-identifying-and-discrimination-risk pre-classifier strips photo, date of birth, gender, marital status, religion and nationality fields from any payload that reaches the language model. Second, the ranker's output is a screening aid for the consultant, never an in-or-out decision — every CV that lands on Atelier is also reviewed manually by the named consultant on the role. Third, we audit log every screen run; the audit log is reviewed by the firm partners monthly.
Atelier is not visible to candidates and is not part of how we communicate with candidates. The consultants' conversations with candidates remain entirely human.