Reported operating use · Recruitment operations
AI candidate sourcing: Blue Castle Agency example
Zapier reports that Blue Castle Agency uses Claude through Zapier MCP to read active jobs and company, role and market profiles, build candidate lists and check them in a second pass. A separate server process prepares submissions, and recruiters handle candidate conversations.
Research candidate lists and prepare recruiter submissions
Blue Castle Agency
Reported recurring sourcing and submission preparation at a recruitment agency; recruiters handle candidate conversations.
A published third-party example, not a Nuvepro customer case or evidence of a partnership.
The workflow at a glance
Candidate research and submission preparation
Illustrative summary of the published account
Reported work components. Their arrangement does not imply a fixed execution order or simultaneous work.
- Context
Active jobs and ATS profiles
Claude reads company, role and market context before using sourcing tools.
- List checking
A later Claude pass checks lists
The same context supports a second-pass check; this alone is not independent verification.
- Submission process
Notes and transcripts against the role
A server process prepares submissions while recruiters conduct candidate conversations.
What the source reports
Zapier reports that Claude reads active jobs and company, role and market profiles from Blue Castle’s applicant tracking context, then uses sourcing tools to build lists. A later Claude pass checks the lists with the same context. A separate server process compares recruiter notes and interview transcripts with the job profile and prepares submissions. Recruiters focus on candidate conversations.
Reported technology: Anthropic Claude and Zapier MCP. Scope and outcomes are source-reported.
What the enterprise environment would need
The following is a proposed project design for an enterprise doing similar work. Published details are marked; the remaining systems, data and scenarios would be agreed for your project.
Proposed practice environment
Prepare an evidence-linked shortlist
Illustrative project design to scope with your enterprise.
- 1 · Set up the environment
Fictional ATS and sourcing records
Keep stable job/candidate IDs, approved criteria and source-linked experience evidence.
- 2 · Preparation and agent contributions
Agent drafts shortlist and submission
Compare profiles, notes and transcripts; preserve ambiguous identity and contradictory evidence.
- 3 · Independent checks
Independent recruiters verify evidence
Inspect incorrect inclusions and exclusions against held-out records and permitted criteria.
- 4 · People decide acceptance
Recruiter accepts, corrects or rejects
Reconcile identity, current requirements and support before accepting a submission packet.
A reviewable evidence package
A shortlist and recruiter submission packet with traceable qualifications and gaps.
If a check fails → withhold acceptance.
Both model passes repeat an unsupported experience claim. Independent review must detect and exclude it.
Failure or owner rejection returns to review
- Hold
- → 2 · Revise work
- → 3 · Rerun independent checks
- → 4 · Owner review
Operating boundary. Research and drafts do not authorize contact, progression or a final hiring decision.
- Systems and tools
- Published: Claude uses existing recruiting tools through Zapier MCP; submission preparation also runs on the agency’s server. Proposed: use approved ATS and sourcing interfaces with test records.
- Data and state
- Proposed: maintain stable job and candidate IDs, approved role criteria and source-linked experience evidence across lists, notes and transcript records.
- People and handoffs
- Published: recruiters conduct candidate conversations. Proposed: reviewers resolve conflicting evidence and decide whether a submission is supportable.
- Rules and authority
- Proposed: separate research, record updates, candidate contact and progression permissions. Generated fit scores must not invent experience or supply authority to contact someone.
- Normal and exception cases
- Proposed: test closed jobs, stale profiles, ambiguous identities, contradictory transcripts and missing qualification evidence.
- Reset and replay
- Proposed: version job criteria and candidate evidence; rerun after a changed role without silently reviving rejected claims.
A proposed simulation
An illustrative practice design derived from this work, not an executed assessment.
Prepare an evidence-linked shortlist for an active role and a submission packet a recruiter can accept, correct or reject.
Review gate: Reconcile candidate identity, current role criteria and supporting evidence; distinguish a proposed fit judgment from a verified fact.
Challenge the result: Give both model passes the same plausible but incorrect experience claim. An independent check must detect the unsupported inclusion.
What needs checking
- Published: a second Claude pass checks sourcing lists; this alone does not establish independent verification.
- Proposed: independent recruiting reviewers qualify the evidence and allowed criteria for held-out records; inspect incorrect inclusions and exclusions separately.
- Proposed: confirm that unresolved identity or unsupported experience reaches review and cannot trigger contact or a final hiring decision.
Review environment fidelity, human capability and agent performance separately. The enterprise’s accountable owners decide acceptance of the shared workflow and any operating use.
Questions about this example
- What does Blue Castle Agency use AI for in sourcing?
- Claude reads active jobs and profiles from the agency's applicant tracking context, then uses sourcing tools to build lists. A later Claude pass checks them, and a separate server process compares recruiter notes and transcripts with the job profile to prepare submissions.
- Who checks the AI's candidate lists at Blue Castle?
- Recruiters conduct candidate conversations. A second Claude pass checks the lists, but that alone does not establish independent verification. As a proposed check, not Blue Castle's, independent recruiting reviewers would qualify the evidence against held-out records.
- How strong is the evidence for Blue Castle's account?
- It is a supplier-published operator account, not an independent hiring-outcome study. Reported approval rates and time savings are operator estimates. It does not establish enterprise-scale validation or the completeness of its data-permission and decision controls.
- What would an enterprise need to build something similar?
- This is Nuvepro's proposed design, not Blue Castle's: approved ATS and sourcing interfaces with test records, stable job and candidate IDs, and versioned criteria and evidence. Unresolved identity or unsupported experience must reach review and never trigger contact or a hiring decision.
Shape this around your enterprise
Bring a similar workflow and the people responsible for it. Together, define the records and systems to use, the result your team will build or configure, and how its owners will check it. Nuvepro scopes the simulation, environment and preparation. Agree any existing-agent evaluation, implementation assistance and production responsibilities separately.
Sources and limitations
Curation review: . This is the date this source example was reviewed, separate from source publication dates.
- This is a supplier-published operator account, not an independent hiring-outcome study. Reported approval rates and time savings are operator estimates. The page does not establish enterprise-scale validation or the completeness of its data-permission and decision controls.
- Zapier: How Blue Castle Agency uses Claude and Zapier MCP to source candidatesSource publication: 2026-10-22
More hr and recruiting examples
- AI recruiting shortlists: Adecco exampleThe Adecco Group
- AI candidate sourcing: Anthropic exampleAnthropic
- AI hiring outreach: Alex Lieberman exampleHiring-outreach workshop example (Alex Lieberman)
- AI HR service desk: IBM exampleIBM