One Pipeline, Seven Stages
Care UK's screening automation is an orchestrated pipeline of seven stages, each handling one part of the candidate journey from raw CV to recruiter-ready profile. Understanding how the stages fit together explains the two properties that matter most in care-sector hiring: consistency and auditability.
The design principle throughout: AI reads and organizes; humans decide. Every stage produces structured, logged output that the next stage — and ultimately a human recruiter — consumes.
Stage 1 — Resume Parsing
Candidate CVs arrive in every imaginable format: polished PDFs, scanned documents, exported word-processor files. DocQ's extraction models parse each one into a structured profile — personal details, qualifications, employment history, certifications — without any model training or labeled datasets. The parser works from day one, which mattered for a deployment measured in weeks rather than quarters.
Stage 2 — Application Pre-fill
The structured profile flows directly into Care UK's application forms on DocQ's forms platform. Candidates confirm rather than re-type; recruiters receive applications whose data already matches the CV. This single step eliminates the double-entry that used to consume recruiter hours and introduce transcription errors.
Stage 3 — Gap Analysis
Employment history is scanned for unexplained gaps — a core safeguarding control in care hiring, where regulators expect providers to account for a candidate's full history. Detected gaps aren't verdicts: they're structured findings, attached to the profile, that the recruiter must explicitly review and resolve.
Stage 4 — ID & Visa Extraction
Identity documents and right-to-work evidence are processed by the same extraction engine: passports, biometric residence permits, visas. Extracted fields are validated against the application data, and discrepancies are surfaced as findings. The originals stay attached to the candidate record — evidence and conclusion in one place.
Stage 5 — Scoring & Red Flags
Every candidate is evaluated against the same role-specific criteria. The scoring model produces a structured assessment — strengths, concerns, and explicit red flags — and each flag carries its reasoning. There is no black-box ranking: a recruiter looking at a flagged profile sees what triggered the flag and the underlying evidence.
Consistency here is the point. The same candidate profile produces the same assessment regardless of which of 120+ homes is hiring or which recruiter picks up the file.
Stage 6 — Screening Questions
Role-specific screening questions are issued through governed DocQ forms, with responses captured as structured data on the candidate record. Conditional logic adapts the question set to the role and the earlier pipeline findings.
Stage 7 — Recruiter Matching
Finally, the pipeline routes each screened candidate to the right recruiter for the right home, based on role type and location. Matching that previously depended on someone knowing who handles what now happens automatically — and the assignment, like everything else, is logged.
The Governance Layer
What makes this architecture suitable for a regulated sector isn't any single stage — it's the substrate. Every stage runs as a step in a DocQ workflow, which means:
- Attribution — every action, human or AI, is recorded with who, what, and when.
- Explainability — AI outputs are structured findings with reasoning, not opaque scores.
- Separation of duties — AI prepares, humans decide, and the workflow enforces the order.
- Reconstructibility — any hiring decision can be replayed end to end from the record.
The pipeline is configuration, not code: criteria, question sets, routing rules, and flag thresholds are all adjustable by the team without redevelopment. When hiring policy changes, the pipeline changes with it — and the audit trail records that, too.
