The five most time-consuming tasks at a Swiss recruitment consultancy are CV screening, dossier production, matching, interview preparation, and client briefing. That is the order to analyse them in, but not necessarily the order to automate them in.
If you want to reduce Time to Hire, first understand where the hours actually go. Most consultancies overestimate sourcing and underestimate dossier production: two hours per candidate for copy-paste into a Word template is the default in small boutiques. This post lays out the five biggest time sinks, shows what is realistic to automate, and ranks the tasks by Swiss FADP (revDSG) and GDPR risk.
Key takeaways
- A typical consultant day is 40-60 % non-billable: screening, formatting, follow-ups, clarifications.
- Dossier production is the hidden 2-hour tax per candidate, and the biggest lever if you want to reduce Time to Hire.
- Automated scoring is legally trickier than dossier generation: GDPR Art. 22 requires "meaningful human involvement".
- Data residency in Switzerland is not a marketing add-on but the operating frame for any AI-assisted recruiting tool.
- Priority order: dossier first, interview prep second, screening and matching carefully, client briefing as a consequence.
Where the week goes: the hidden hour budget
The SHRM Talent Acquisition Benchmarking Report (2022) puts average time-to-fill at 36 days. The LinkedIn "Future of Recruiting" 2024 study shows recruiters spend roughly a quarter of the week on administrative work, not consulting, not sourcing, not interviews.
For a Swiss boutique with three to five consultants, a typical 45-hour week looks roughly like this:
| Task | Share | Per week (h) | Billable? |
|---|---|---|---|
| Sourcing and active search | 15 % | 6.75 | Partially |
| CV screening / shortlist | 20 % | 9.00 | No |
| Dossier production | 20 % | 9.00 | No |
| Interview preparation | 10 % | 4.50 | Partially |
| Interviews (delivery) | 15 % | 6.75 | Yes |
| Client briefing / hand-off | 10 % | 4.50 | Yes |
| General admin | 10 % | 4.50 | No |
The numbers are industry estimates, not Wield cohort data. But the pattern holds: more than half the week is formatting, checking, copying. To reduce Time to Hire, go after that non-billable block, not the two hours of interview time.
Task 1: CV screening and shortlisting
Why it takes so long
An IT role in Zurich typically pulls 80 to 200 CVs into the inbox. Gartner research on recruiter productivity (2023) shows screening consumes 30 to 40 % of sourcing hours. In Switzerland the problem is sharper because of bilingual applications (dossier in DE, CV in FR), a heterogeneous qualification landscape (FH, university, eidg. Fachausweis, Berufsprüfung), and lateral entrants from the EU with unfamiliar credential formats.
What can be automated
Structured field extraction from PDF and DOCX is a solved problem today. Embedding-based matching between job spec and CV produces a sensible pre-sort, not as a final decision, but as a filter layer. The consultant sees fifteen candidates instead of two hundred, without a machine rejecting anyone.
Where it breaks
Handwritten CVs from hospitality or construction are not reliably read by any parser. If more than 30 % of your inbound is handwritten, extraction tools save little. The work stays manual here.
Task 2: dossier production, the hidden 2-hour tax
Why it takes so long
Dossier production is the most underestimated time sink in consultancy work. A typical dossier for a Swiss client includes:
- A structured table of contents with the career timeline.
- Anonymisation (name, photo, current employer) for the blind dossier.
- Bilingual passages when the client is in Geneva and the candidate works in Zurich.
- Client branding: logo, colour palette, template variant per client.
- Photo correction when the candidate's headshot arrives unusable.
60 to 120 minutes per candidate is realistic. For a mandate with four shortlisted candidates that is roughly half a working day, and the quality depends on whoever touched the file last.
What can be automated
Schema-first template fillers generate a consistent dossier layout from the structured fields. Anonymisation becomes a toggle, not manual work. Bundling several candidates into one share link replaces the email attachment flow.
Where it breaks
A completely new client with a bespoke template still requires the first manual pass. Automated generation only helps once the template is set up. For one-shot mandates with an exotic layout, manual is often faster.
A concrete scenario: bilingual client, four shortlisted candidates
A Geneva client asks for a dossier in French. Three of the four shortlisted CVs are in German, the fourth in English. Classic flow: translate every passage, drop into the client template, anonymise each file separately, export to PDF, email with a password. 90 to 120 minutes per dossier, plus half an hour for the bundle. Four dossiers = roughly eight hours of work for a document the client will read in ten minutes. Set up the template stage once and that block collapses to review time, typically 15 to 20 minutes per candidate.
Task 3: matching and scoring
Why it takes so long
Mental scoring lives in the consultant's head. Each person sorts differently, cross-team calibration is missing. At the debrief you argue whether "fit" means the same thing. Outcome: candidates drop out whom someone else would have kept in.
What can be automated
Cosine similarity between JD and CV embeddings plus structured field matching (skills, seniority, languages, region) delivers a traceable ranking. Important: the ranking explains what is similar, not why the client wants this person. That remains the consulting work.
Where it breaks
Scoring without a human decision is risky under GDPR Art. 22 and revDSG Art. 21: a solely automated decision with legal effect is not permitted. Any matching tool has to keep the consultant as the decision-maker. Skip that and a subject access request becomes a real problem.
Task 4: interview preparation
Why it takes so long
Every role calls for slightly different questions. Senior interviews need different depth than junior ones. Debriefs cannot be compared when each consultant runs a different structure. Outcome: gut decisions instead of calibration.
What can be automated
Structured interview protocols generated from job spec plus shortlisted CV, question bank, scoring rubric, follow-up points. The consistency makes debriefs comparable. A consultant can be swapped without the process tipping over.
Where it breaks
Executive search and C-level mandates are too bespoke for template generation. Consulting experience carries the interview, not the protocol. For mid-seniority and below, the approach works.
Task 5: client briefing and shortlist hand-off
Why it takes so long
At the end of every shortlist sits the hand-off: cover letter, comparison table, short profiles, dossier bundle. Every client wants it slightly differently. Formatting routinely takes longer than the writing.
What can be automated
A bundled dossier set with a public share link replaces the email with five attachments. An auto-generated comparison table puts the shortlist side by side in one view. The client sees exactly the passage that matters, not a 40-page collection PDF.
Where it breaks
If the client wants a live walk-through, the document is just preparation material. The conversation is still the work. Automation does not replace the presentation, it replaces the production behind it.
Reduce Time to Hire under revDSG and GDPR: where to start
Not every automation pays out equally, and not every one carries the same compliance profile. An honest priority list looks like this:
| Task | Automation effort | Time recovered | Art. 22 risk |
|---|---|---|---|
| 2. Dossier | Medium | Very high | Low |
| 4. Interview protocol | Low | High | Low |
| 5. Client briefing | Low | Medium | Low |
| 1. CV screening | High | High | Medium |
| 3. Matching / scoring | High | High | High |
Start with Task 2 (dossier). The recovered time is the cleanest and the compliance risk is minimal: formatting a document is not a legal decision.
Task 4 (interview prep) follows. Quick to implement, little legal risk, and the quality gain from calibration is real.
Tasks 1 (screening) and 3 (matching) last, and only with a human gate. Sliding from pre-filtering into deciding without "meaningful human involvement" under GDPR Art. 22 risks more than the time it saves.
On top of that sits data residency: revDSG Art. 6 requires adequate protection of personal data. If you run candidate data through a US-hosted AI, you have to document the processor, the sub-processors, and the transfer basis. Starting on Swiss infrastructure is simpler. That means Swiss hosting and AI inference in a Swiss cloud region. Details on the Wield architecture at /#trust.
How to measure Time to Hire in your own week
Before buying any tool, spend a week measuring. Three steps:
- Log hours by task type. For one week, assign every 30-minute block to one of seven categories (sourcing, screening, dossier, interview prep, interview, briefing, admin). Excel is enough.
- Billable check. Mark each block as covered by the fee or not. The non-billable share is the target.
- Set priorities. The task with the highest non-billable share and the lowest Art. 22 risk gets automated first. In most Swiss boutiques that is Task 2, not Task 1.
The exercise takes one week and shows where the money actually goes. No tool-evaluation process replaces this step.
Where these fixes don't land
Every automation has a customer segment where it does not carry:
- Executive search. A CEO mandate is hand-curated by definition. Templates help the assistant, not the partner.
- High share of handwritten CVs. Construction, hospitality, parts of healthcare still deliver paper regularly. Extraction tools top out around 70 % here; the rest stays manual.
- Regulated sectors with on-prem requirements. If the client requires candidate data to stay in their own infrastructure only, cloud SaaS usually drops out.
- One-shot mandates with exotic client templates. The first dossier in a brand-new layout is faster by hand. Setup pays off from the second one onward.
If you sit in one of those four camps, keep automation to the non-critical tasks and stay honest about the critical ones.
What Wield ships that helps
Wield covers the five tasks from a Swiss workspace: extraction from PDF, DOCX and scanned documents, JD-to-CV matching with embeddings, schema-first dossier generation including anonymisation, structured interview protocols, and bundled share links for the client briefing. Hosting runs at Infomaniak in Geneva, AI inference on Vertex AI in the Europe-West6 region (Zurich). Audit trail, tenant isolation, and deletion workflows are wired end to end.
Product overview: /#product. Pricing for boutique and team sizes: /#pricing.
Closing
To reduce Time to Hire is not to replace the consulting work. It is to take the ten non-billable hours per week out of the workflow that nobody pays for as consulting. Start with Task 2 and work upward from there, and you recover time without picking up a GDPR problem.
For questions or a structured look at your own workflow, reach us at info@tecminds.ch or directly via /#pricing.