From 18 days to 4: better shortlists in a fraction of the time.
A 62-person executive search firm specializing in CFO, COO, and CHRO mandates deployed six production agents on OpenClaw — running entirely on the firm's own infrastructure — and trained 19 consultants and researchers to extend them. Time-to-shortlist fell from 18 days to 4, placements per consultant rose 2.2×, and the firm added €1.2M in placement fees in eight months.
The client is a Central European executive search firm specializing in senior finance, operations, and HR mandates: CFOs, COOs, CHROs, and the divisional equivalents inside large enterprises. Sixty-two people across three offices, founded in 2009, a respected brand in mid-cap and private-equity-backed companies across the region. They typically hold 80–120 active mandates at any given time.
Their problem was structural, not technological.
Executive search has not really changed since the 1980s. The work is research-intensive and judgment-intensive: a consultant takes a client briefing, a researcher maps the universe of candidates (typically 200–600 named people for a senior mandate), the consultant qualifies them through phone calls and references, the firm produces a shortlist of 5–8 candidates, and then the slow choreography of interviews and offer-management begins. The bottleneck has always been the research and qualification stage. A senior researcher can map perhaps 80–120 candidates per week with reasonable depth. A typical mandate requires 3–4 weeks of research before the consultant can even begin to qualify.
The firm was losing mandates because they were too slow. PE-backed clients in particular were running parallel beauty contests with two or three search firms; whichever firm produced credible shortlists first usually won the relationship for the placement and several follow-on mandates. The firm's win rate on PE pursuits had dropped from 41% in 2021 to 27% in 2024 — almost entirely because they were arriving with strong shortlists 10 days after their faster competitors.
They engaged Orchestrary with two non-negotiable constraints:
- Confidentiality. The candidate database, the active mandates, the reference notes, the interview transcripts — all of it is some of the most sensitive data in the professional services industry. Pasting any of it into a hosted commercial AI service would have been a contract violation in approximately 100% of the firm's mandates.
- Quality of judgment must not degrade. Executive search is a reputation business. A bad shortlist is far worse than a slow one.
Both constraints pointed to OpenClaw, deployed in the firm's own cloud tenant, with senior consultants kept tightly in the loop on every shortlist decision.
The constraint: a search firm cannot afford to lose its judgment
Executive search firms are particularly vulnerable to a specific kind of AI failure: a system that produces plausible-looking shortlists by pattern-matching on superficial criteria (job titles, company logos, years of experience) while missing everything that actually matters in senior placements (executive presence, situational fit, transformation experience, board readiness, cultural alignment with a specific PE sponsor's style). A firm that ships such shortlists to clients destroys its brand within six months.
This shaped the entire engagement. Every agent was designed around a discipline we adopted on day one: the agent expands the surface; the consultant makes the call. No agent in this firm produces a shortlist autonomously. No agent makes a placement recommendation.
The portfolio: six agents covering the search lifecycle
Wave 1 — Built by Orchestrary in months 1–4
MANDATE-BRIEFER · Client briefing to structured search brief
- Time per brief: 4–6 hours → 45–60 minutes
- Brief completeness: 78% → 96% of items addressed in first draft
- Client revision requests: 1.4 per mandate → 0.3
TALENT-MAPPER · Candidate universe mapping
- Time per universe map: 3–4 weeks → 3–4 days
- Universe size: ~200 → ~600 candidates (deeper into the long tail)
- Researcher time per mandate: 120+ hours → 18–24 hours
PROFILE-COMPOSER · Candidate profile drafting
- Time per profile: 3–5 hours → 30–45 minutes
- Total profiles per mandate: ~35 → ~80
- Profile depth improved (surfaces public material that researchers under time pressure routinely missed)
REFERENCE-SYNTHESIZER · Reference call synthesis and pattern detection
- Time per synthesis: 3–4 hours → 30 minutes
- Pattern detection: flags subtle reference patterns that consultants under time pressure routinely missed
- Reference calls per mandate: ~20 → ~30 (firm could afford the larger reference cost)
Wave 2 — Built by the firm's own team using OpenClaw (months 5–8)
LONGLIST-QUALIFIER · PROPOSAL-GENERATOR
LONGLIST-QUALIFIER (built by a consultant and senior researcher) performs initial qualification on the long list — assessing each candidate against must-haves and nice-to-haves, flagging the strongest 60–80 for human consultant review. Reduced longlist qualification from ~4 days to ~8 hours per mandate. PROPOSAL-GENERATOR (built by the BD director) drafts pursuit proposals from the prospect's published context, the firm's relevant prior placements, and proposed team biographies. Reduced proposal drafting from 1–2 days to ~4 hours per pursuit.
The aggregate impact: a search firm that compounds against itself
| Metric | 12 months prior | 12 months from start (annualized) | Change |
|---|---|---|---|
| Active mandates simultaneously | ~95 | ~148 | +56% |
| Time-to-shortlist (median) | 18 days | 4 days | −78% |
| Placements/consultant/year | 4.6 | 10.1 | +120% |
| PE pursuit win rate | 27% | 41% | +52% |
| Mandate-to-placement | 71% | 82% | +15% |
| Placement fees (12-mo rolling) | €11.4M | €12.1M | +€1.2M |
| Client repeat-mandate rate | 38% | 58% | +53% |
Consultant time on research operations: ~35% → ~9%. Consultant time on client-facing work: ~40% → ~64%. Researcher capacity: 3–4 active mandates → 10–14 active mandates per FTE. Universe coverage: ~200 → ~560 candidates per mandate.
Strategic outcome
The firm has restored its PE win rate to a firm-historical high. They are now the firm clients call when they need a shortlist on a compressed timeline — a category that did not exist as a competitive position before this engagement. Two of the firm's largest PE clients have made the firm their default search partner across the portfolio.
Why "deploy AND teach with OpenClaw" was right for executive search
1. Candidate confidentiality is the entire business
Senior executives in active or quasi-active employment do not appear on candidate lists at hosted commercial AI services without their knowledge being a significant breach. The firm's General Counsel made this absolutely clear: any system processing candidate names had to run on the firm's own infrastructure, with no telemetry leaving the boundary, with auditable access logs, and with a kill switch the firm could throw without an external vendor's involvement.
2. The firm's competitive moat is its proprietary data
The firm has 16 years of accumulated mandate notes, candidate qualification records, reference notes, and placement outcomes. OpenClaw runs against the firm's data on the firm's machines. The data does not move.
3. Senior consultants must remain in the loop
Every shortlist decision, every placement recommendation, every reference call interpretation requires human judgment from a person whose name is on the placement guarantee. OpenClaw's terminal-native, file-based design fits this discipline naturally.
The Academy — adapted for a search firm
The two Track 3 graduates who lead the firm's "Agent Engineering" function each spend ~30% of their time on the agent stack and the remainder on their normal work. They are the highest-leverage staff in the firm.
The human dimension
"The single most important number in this engagement is our PE win rate. It is back where it was when we were hungrier and faster, and it is staying there. We did not get faster by cutting corners — we got faster by building a research operation that runs at a different speed than anyone else in our market."
Managing Partner
"I was the most skeptical person in the building when this started. The thing that turned me around was watching TALENT-MAPPER surface a candidate who turned out to be the placement on a mandate I was leading. I would not have found her in a month of manual mapping. The agent does not replace what I do. It expands what I can see."
Senior Consultant · 15 years at the firm
"My team used to spend 80% of their time on the parts of research nobody enjoys — systematic universe mapping, LinkedIn scraping, profile drafting. Now they spend that time on the parts they actually trained for: qualitative qualification, network conversations, judgment. They are happier and they are doing better work."
Head of Research
"I have signed off on every aspect of this system. The combination of on-premise OpenClaw deployment, full audit logging, and the firm's existing access controls means I can defend this to clients, candidates, regulators, and our own insurance providers."
General Counsel
What this engagement model does NOT do
- It does not build a "candidate matching algorithm" that ranks people on a 0–100 score and presents that to clients. The firm explicitly rejected this. Every shortlist is a senior consultant's judgment.
- It does not eliminate the researcher role — it transforms it. The firm's researchers do qualitative work now, not data-entry work, and the firm hires them differently.
- It does not turn a search firm into a software company. The firm runs on agents; it is not selling them.
The deliverable
The contracted commitment was eight months. We exited on schedule. The firm runs the agents themselves, has built two more we never touched, and is hiring its next generation of researchers against a job description that did not exist when we started.
"We did not buy software. We built a different version of our firm. The fact that you can stop coming here is the proof that we got what we paid for."
Managing Partner · Engagement closeout
From outbound to outcomes: 1.9× meetings, 27% lift on close rates, no new sales hires
Want to be the seventh case study?
A 60-minute discovery call. No software pitch. We map your most painful workflow, scope a first agent, and tell you honestly whether this engagement model fits your organization.