3 ready-to-build workflows

AI agent workflows for HR & recruiting

Kill the manual screening, scheduling and onboarding paperwork.

Each recipe below is expressed only in VegaDūta's real workflow building blocks — Trigger, Agent, Knowledge, Condition, Approval, Output, Tool (MCP), Voice, Device, Loop, Code and Parallel — so it maps 1:1 to something you can assemble in the Workflow designer. Consequential actions always pass a human Approval step.

1. Résumé screening & shortlisting (manual → minutes)

A recruiter opens every CV by hand, reads it against the JD, and copies notes into a sheet — hours per role, and the 200th CV gets less attention than the first.

Trigger: Webhook (new application) or Email (careers inbox)

How the workflow runs

  1. Trigger

    Application lands. A Webhook (from your ATS/careers form) or an Email trigger on the careers inbox starts the flow per candidate.

  2. Knowledge

    Score against the JD. A Knowledge node holds the job description + must-haves; the résumé is compared against it, not a keyword guess.

  3. Agent

    Summarize + rate. An Agent node writes a 3-line summary, a fit score, and the gaps — consistently, for every candidate.

  4. Condition

    Auto-route by fit. A Condition node splits strong / maybe / reject so recruiters only read the shortlist.

  5. Approval

    Recruiter confirms shortlist. An Approval node keeps the human as the decision-maker — the agent ranks, a person decides.

  6. Tool (MCP)

    Write back to the ATS. A Tool (MCP) node updates the candidate stage and notes in your ATS.

Channels & connectors

  • Webhook
  • Email
  • Knowledge base
  • Approval / HITL
  • Tool (MCP → ATS)

Outcome

Every applicant is read and scored the same way within minutes; recruiters spend their time only on the shortlist and the human decision.

Why it helps

Removes the fatigue bias of manual screening and the copy-paste into sheets, while the hire/reject call stays with a person.

Build spec

1 agent — 1 screening agent per applicant; routing + ATS write-back are workflow nodes.

System prompt (paste-ready)

You are a recruiting screener. Given a résumé and the job's must-haves (from Knowledge), output strict JSON {summary (max 3 lines), fit_score 0-100, matched[], gaps[], recommendation: STRONG|MAYBE|REJECT}. Score only on job-relevant evidence in the résumé. Do NOT use name, gender, age, religion, caste, or photo as signals — ignore them entirely. If a must-have can't be confirmed, list it in gaps, don't assume it.

MCP connectors

  • ATS (Greenhouse/Lever/etc.) — via a Tool (MCP) node
  • careers inbox — via Email trigger

Built-in tools

  • knowledge_search (JD + must-haves)
  • file parsing (résumé PDF/DOCX)

Guardrails

  • Bias guardrail: protected attributes are explicitly excluded from scoring in the prompt
  • The agent ranks; it never rejects — the Approval node (recruiter) makes every advance/reject decision
  • Same rubric for every candidate — deterministic Knowledge-grounded criteria, not vibes

Output & delivery

Per-candidate JSON → Condition buckets STRONG/MAYBE/REJECT → recruiter reviews the shortlist at the Approval node → a Tool node writes the stage + notes back to the ATS via MCP.

2. Interview scheduling without the email ping-pong

Coordinators trade a dozen emails to find a slot across a candidate and 3 interviewers.

Trigger: Webhook (candidate advanced to interview)

How the workflow runs

  1. Trigger

    Candidate moves to interview. A Webhook trigger fires when the ATS stage changes.

  2. Agent

    Propose times. An Agent node reads panel availability and drafts options.

  3. Output

    Offer slots on WhatsApp/Email. An Output node lets the candidate pick in one tap.

  4. Tool (MCP)

    Book it everywhere. A Tool (MCP) node writes the invite to the panel's calendars.

Channels & connectors

  • Webhook
  • WhatsApp
  • Email
  • Tool (MCP → calendar)

Outcome

A confirmed slot on everyone's calendar with zero back-and-forth.

Why it helps

Collapses days of coordination into a single self-serve pick.

Build spec

1 agent — 1 scheduling agent; the calendar write is a workflow Tool node.

System prompt (paste-ready)

You are an interview-scheduling coordinator. Given a candidate advanced to interview and the panel {interviewers[], durations, timezone}, read each interviewer's availability and propose 3–4 non-conflicting slots that fit everyone. Offer them to the candidate to pick in one tap. Only propose times that are actually free on every required panel member's calendar — never double-book. Confirm the final choice back to all parties.

MCP connectors

  • Calendar (Google/Microsoft 365) — via a Tool (MCP) node
  • ATS — inbound via Webhook trigger
  • WhatsApp / Email — candidate channel

Built-in tools

  • http_request (read panel availability)

Guardrails

  • Proposed slots must be free on every required panel member's calendar — no double-booking
  • The invite is written by the Tool node only after the candidate picks — no speculative holds
  • The timezone is respected for both candidate and panel

Output & delivery

The agent proposes conflict-free slots, the candidate picks one on WhatsApp/Email, and a Tool node writes the confirmed invite to every panel member's calendar.

3. New-hire onboarding checklist that runs itself

IT access, documents, and welcome steps are chased manually across three teams for every joiner.

Trigger: Webhook (offer accepted)

How the workflow runs

  1. Trigger

    Offer accepted. A Webhook trigger starts onboarding the moment HR marks the hire.

  2. Parallel

    Kick off every track at once. A Parallel node fans out to IT provisioning, document collection and payroll setup simultaneously.

  3. Tool (MCP)

    Provision + request docs. Tool (MCP) nodes raise the IT tickets and pull the paperwork tasks in your systems.

  4. Output

    Guide the joiner. An Output node walks the new hire through each step on WhatsApp/Email and nudges what's pending.

Channels & connectors

  • Webhook
  • Parallel
  • Tool (MCP → IT/HRIS)
  • WhatsApp
  • Email

Outcome

Day-one readiness is tracked automatically; nothing falls through the cracks between teams.

Why it helps

Replaces the manual chase across IT/HR/payroll with a single orchestrated flow.

Build spec

1 agent — 1 concierge agent guides the joiner; the Parallel fan-out and the provisioning Tool nodes are workflow steps.

System prompt (paste-ready)

You are a new-hire onboarding concierge for {{company}}. Given a joiner {name, role, start_date, tracks[]}, guide them through each onboarding step (IT access, documents, payroll, welcome) in a friendly WhatsApp/Email conversation. Track what's done vs pending across all tracks and send timely nudges for anything outstanding before the start date. Answer common joiner questions from the onboarding Knowledge base. Never ask for sensitive documents over an insecure channel — link to the secure upload instead.

MCP connectors

  • IT provisioning / ITSM — via a Tool (MCP) node
  • HRIS / payroll — via a Tool (MCP) node
  • WhatsApp / Email — joiner channel

Built-in tools

  • knowledge_search (onboarding policies/FAQs)

Guardrails

  • Sensitive documents are collected only via the secure upload link, never over chat
  • Provisioning tickets are raised by Tool nodes in your systems — the agent tracks status, it never grants access itself
  • Each track's status is tracked so nothing is skipped between IT, HR and payroll

Output & delivery

A Parallel node fans out IT/document/payroll tracks; Tool nodes raise the tickets and pull the tasks, while the agent guides the joiner step-by-step on WhatsApp/Email and nudges whatever's still pending.

Related industries

Build one of these in minutes

The sandbox gives you a live agent workspace — no account, no card. Or head back to the full catalogue and compare patterns across every industry.

Try this in the sandbox — no signupBrowse all industries
See how VegaDūta compares to n8n, Dify and BotpressWhy VegaDūta's architectureEstimate your WhatsApp API costs