Recruiting Sourcing: Candidate Research and Interview Scheduling
💡 One-sentence summary: Use a Grok Bot to build your AI recruiting sourcing — source candidates against the job description, leave evidence for every match, draft outreach in your tone of voice, and schedule interviews via your calendar; contacting candidates always stays behind approval.
This is lesson four on Route A (workplace), using the same official six-step method as Daily Briefing, Expense Reconciliation, and Sales Outreach. This lesson corresponds to the official use-cases page, where it is one of the 8 roles: the Talent scout (talent acquisition specialist). The official positioning for it: owns sourcing, candidate research, and outreach drafts — not "helping you write a few recruiting DMs," but a candidate list where the evidence for every match is clearly explained, and contacting candidates always stops at a review point. The four types of connections officially named are ATS, approved sourcing tools, email, and calendar — the calendar is where the "interview scheduling" in this lesson's title comes from.
The whole lesson takes about 10 minutes. The officials also have one special requirement for this role that other roles don't have: the Bot must respect candidate privacy and the terms of each source; on top of that comes the "outreach stays behind approval" rule common to all roles — step 6 of this lesson is dedicated to how to put both into practice.
What you'll be able to do after this lesson
- Build a "recruiting sourcer" Bot: connect your ATS, sourcing tools, email inbox, and calendar, and source candidates against the must-have criteria in the job description
- Have the Bot exclude everyone already in your ATS and explain the evidence for every match — which must-have criterion was hit, and where the evidence comes from
- Draft personalized outreach for your selected candidates in your tone of voice; at the interview stage, check the calendar for open slots and draft invitations — everything stays in draft
- Save the sourcing workflow as a skill so you can rerun it with a different job description; every step of contacting candidates stays behind approval
Why a Grok Bot instead of web-based AI?
| Capability | Web-based AI | Grok Bot |
|---|---|---|
| Can read your ATS | ❌ Export the candidate list and copy-paste it into the chat | ✅ Connects directly via connectors; for ATSs without a connector, log in through the cloud PC browser |
| Remembers must-have criteria and your tone of voice | ❌ Re-paste the job description and sample emails in every conversation | ✅ Hand the job's must-have criteria and writing samples to the Bot; context accumulates across conversations |
| Explains the evidence for every match | ❌ Says "great fit" based on the model's gut feeling | ✅ The official starter prompt explicitly requires explaining the evidence for each match |
| Excludes people already in the ATS | ❌ You compare the lists by eye yourself | ✅ The official starter prompt explicitly requires excluding everyone already in the ATS |
| Can see your calendar | ❌ Scheduling interviews means flipping through the calendar yourself for slots | ✅ After connecting Google Calendar, it can read your calendar, check open slots, and create and update events |
Your current struggle
- The job has been posted for two weeks with hundreds of resumes piled up, yet not a single truly suitable passive candidate has applied
- Browsing sourcing platforms profile by profile, you can only research a handful of people in an evening, and you still can't articulate "why you think they're a fit"
- Once an interview is finally set, back-and-forth emails to find a slot that works for both sides drag on for another week
When to use this technique
When you need to:
- Source a batch of candidates against the must-have criteria in a job description, with evidence attached for every match
- Compress "browsing platforms to do homework" into "reading one evidence-backed candidate list," then approve outreach item by item
- Interview invitations that automatically include available time slots, with calendar changes and email sending confirmed by you
And you don't want to:
- Manually browse sourcing platforms and hunt down people who are already in the ATS
- Mass-send templated recruiting DMs, or let AI grab candidate profiles at random
🎒 Preparation Before Starting
- Read 3.2 Skills and teaching by demonstration — step 4 of this lesson saves the sourcing workflow as a skill, and this is also the shared prerequisite for all Route A scenarios
- Have a Bot you can assign work to: if you don't have one yet, first go to 1.5 Create Your First Bot
- Think through the four data sources: ATS (connector, or web URL + login method), sourcing tools (your approved sourcing channels), outreach email (Gmail or Outlook), and calendar (Google Calendar)
- Prepare two materials: a job description with clearly stated must-have criteria + 2–3 of your best-written candidate outreach emails as writing samples
- About 10 minutes, plus one round of human feedback after the first list comes out
Follow Along
Step 1: Write the recruiting sourcer's responsibilities into the Bot's description
Why: The first step of the official six-step method is to write the responsibilities, data sources, output format, and standing boundaries into the Bot's description — it relies on this text every day to understand its job. The privacy and source-terms requirements specific to Talent scout should be in there from day one.
How to do it: Create or open a Bot (sidebar + → Create a Bot), and write the description like the following, adapted to your actual situation:
You are my recruiting sourcer, responsible for candidate sourcing, background research, and outreach drafts.
Data sources: my ATS, approved sourcing tools, outreach email and calendar, plus the job descriptions and writing style samples I provide.
Output format: a candidate list (each entry with matching evidence) + personalized outreach drafts written in my voice + interview invitation drafts.
Standing boundaries: only source, research, and draft—do not contact any candidates, do not modify records in the ATS; skip anyone already in the ATS;
Respect candidate privacy, use only the sourcing channels I have approved, and comply with the terms of use of each source; state clearly when information cannot be found—do not fabricate it.What you should see: after the description is saved, this responsibility section is visible on the Bot's detail page. From then on, no matter which conversation you assign work in, it operates according to this positioning.
Step 2: Connect the ATS, sourcing tools, email, and calendar
Why: the facts in its research can only come from the data sources you connect—channels that Grok Bot cannot read never make it into the candidate list. The four categories of connections the official guide names for this role are: ATS, approved sourcing tools, email, and calendar.
How to do it: handle each data source one by one; for details of the 5-step connection method, see 2.5 Connect apps:
- Outreach email: search for Gmail (or Outlook) in Connect apps and add it; when the Bot asks for Authorize / Authenticate, complete the login in the browser
- Calendar: search for Google Calendar in Connect apps and add it—the official guide explicitly states that after connecting, the Bot can read the calendars you can see, create and update events, check free time slots, and respond to invitations on calendars you can edit
- ATS: if it's searchable in Connect apps, connect it directly; if there is no connector, have the Bot log into the web version using the browser on its cloud computer—login credentials are stored in Vault, and the Bot logs in on your behalf but cannot see the values; if your ATS offers an MCP server, you can also have the Bot add it by URL
- Sourcing tools: the official guide does not name specific tools, only "approved sourcing tools"—connect the channels you already use whose terms permit sourcing, or give it the web version's address
- If authorization breaks, click Reconnect (on the app) or Reopen (on the connection card in the chat) to get back to the login page
What you should see: email and calendar show Installed; the ATS can be opened in the Bot's browser and stay logged in. Note that Added ≠ logged in—when it shows Needs auth or Disconnected, you must click Authenticate (on mobile it's Authorize) to re-authorize for it to count.
💡 This is the unique value of the Grok Bot
A web-based AI can only see the exported list you paste to it; the Grok Bot has its own cloud computer and browser, so it can log into the web version of an ATS without a connector and look things up itself. There is also a free research channel: the Bot can search and read public content on X without connecting an X account (read-only, not logged in as you)—candidates' public posts and public discussions about a field are all ready-made background research material.
Step 3: Run the first AI recruiting sourcing session with the official starter prompt
Why: the official starter prompt deliberately stops at a read-only, no-modification review point—validate the list quality first, then talk about automation.
How to do it: send the job description and writing style samples to the Bot along with a message (drag them into the input box as attachments), then send this localized version of the official Talent scout starter prompt:
Based on this job description, find 20 candidates who meet the hard requirements.
Exclude everyone already in our ATS, explain the evidence for each match,
and draft personalized outreach in my voice. Do not contact anyone.If you only want it to look at a certain app, type @ in the chat to attach the app to this task.
What you should see: the Bot checks the data sources one by one and returns three things—a list of 20 candidates (with evidence explained for each match), people already in the ATS excluded, and personalized outreach drafts written in your voice; throughout, it contacts no candidates and touches no records in the ATS.
Step 4: Revise the output until it's reviewable, then save it as a skill
Why: the official six-step method requires revising the results until they are "reviewable without you watching," then saving the process as a skill and testing it with a second input—for sourcing, the reviewable standard is: evidence is clearly explained, exclusions are done correctly, and the outreach sounds like you.
How to do it: check the Bot's output item by item; send back anything that doesn't pass for rework (using the "matching criteria feedback" magic prompt below saves effort), and once it's reviewable, solidify the process:
Save the complete candidate sourcing process we just ran as a skill named "Job Sourcing Research."
Requirements: source candidates based on the hard requirements in the job description; exclude people already in the ATS;
explain the evidence for each match; draft personalized outreach based on the writing style samples;
return a reviewable list, contact no one, and modify no ATS records.What you should see: the Bot confirms the skill is saved. From then on, ask for it by name in any conversation—just say "Run the Job Sourcing Research skill" to invoke it (@ is for Bots, groups, routines, and apps; it doesn't bring up skills). The official guide requires testing with a second input—run it again with a different job description and check whether the evidence and writing style remain consistent.
The private skill library is shared across all Bots, so your other Bots can use it too. For saving details, see 3.2 Skills and teaching by demonstration.
Step 5: Schedule interviews—check the calendar for time slots, keep invitations as drafts
Why: the fourth category of connection the official guide names for Talent scout is the calendar. Interview scheduling is the natural next step after sourcing, but every step of it stays in a reviewable state—invitations come out as drafts first, and calendar changes go through your approval.
How to do it: pick the candidates to interview from the reviewable list and send the names to the Bot:
From the reviewable list just now, I've selected these candidates to move on to interviews: [candidate names].
Read my Google Calendar, find 3 45-minute time slots next week when both [interviewer] and I are free,
and draft an interview invitation email for each candidate: explain the role, offer these 3 optional time slots, and ask them to reply to confirm.
Do not send any emails, and do not create or modify any events on the calendar.Note: what the Bot can read is the calendars you can see—the interviewer's calendar can only be checked for free time together if it is visible to you (or shared with the connected account); if you can't see it, check only your own time slots for now.
What you should see: the Bot reads the calendar, checks free time slots, and returns one invitation draft per candidate, each with 3 optional time slots; no emails are sent and the calendar is untouched. After you approve a draft and the candidate replies to confirm a time slot, have the Bot create the interview event on the calendar—it can do this (the official guide explicitly states that after connecting Google Calendar it can create and update events), but as an external action it goes through Auto Review first, and an approval card appears first when your confirmation is needed.
One official boundary to keep in mind: the Google Calendar connector cannot change Meet settings (waiting room, host controls, and the like), nor can it add Zoom links to invitations. The official workaround: fill in the video conference link yourself in the event's location or description field, or create the event in the calendar yourself first and have the Bot update the remaining fields.
Step 6: Hold the approval line—keep outreach behind approval, write privacy and terms into the boundaries
Why: the official guide's closing requirements for Talent scout are two: keep external outreach behind approval, and have the Bot respect candidate privacy and the terms of each source. The first half is a mechanism; the second half you must write into the Bot's boundaries.
How to do it: Mechanically, you don't need to do anything—just keep the defaults. Auto Review is on by default, and outbound emails start as a draft card by default (Needs approval); emails sent by routines are always saved as drafts waiting for you to click Send, and the official docs explicitly state there is no unattended-sending option. On the boundaries side, confirm two things: the description in step 1 already says "respect candidate privacy, only use approved sourcing channels, and comply with source terms"; first-contact emails are one per person with a one-click unsubscribe—if anyone unsubscribes or marks the email as spam, the Bot stops sending to that address and tells you who it skipped.
What you should see: You receive an evidence-backed candidate list, outreach drafts, and invitation drafts, with every step of contacting candidates waiting for your go-ahead.
One easily-missed detail: approval cards triggered by routines expire if not handled within 30 minutes (those initiated in chat wait indefinitely)—so don't add approval-requiring actions like "send invitations" to a sourcing routine; keep it at the list-and-drafts stage. For details on approval rules, see 3.1 Approvals and Auto Review.
📋 Magic Prompt
📝 Candidate Sourcing List (official starter prompt structure)
Expected result: The Bot returns a reviewable list—20 candidates meeting the hard requirements, each match with evidence, people already in the ATS excluded, plus personalized outreach drafts written in your tone.
## Role
You are my recruiting sourcer, responsible for candidate sourcing, background research, and outreach drafts.
## Task
Source 20 candidates based on the hard requirements in the job description, exclude people already in the ATS, explain the evidence for each match, and draft personalized outreach in my tone.
## Input Information
### Required Fields
- Job description: [Full job description including must-have hard requirements]
- Writing style samples: [2–3 of your best candidate outreach emails]
### Optional Fields
- Sourcing channels: [Which approved channels only should be searched?]
- Exclusion rules: [Besides the ATS, who else should be excluded?]
## Output Specification
Reviewable list: candidates ranked by match level, each entry with evidence notes (which hard requirement was hit, where the evidence comes from) + personalized outreach, one draft per person.
## Constraints
- ✅ Exclude everyone already in the ATS; evidence must be traceable back to the source
- ✅ State clearly when information can't be found; do not fabricate
- ❌ Do not contact any candidates
- ❌ Do not modify any records in the ATS📝 Interview Invitation Draft (with calendar time slots)
Expected result: For the candidates you selected from the list, one invitation draft per person, with real free time slots pulled from your calendar—nothing is sent, and the calendar is untouched.
## Role
You are my recruiting sourcer; I've just selected the interview shortlist from your reviewable list.
## Task
Read my Google Calendar to find free time slots and draft an interview invitation email for each candidate.
## Input Information
### Required Fields
- Interview shortlist: [Selected candidates]
- Interviewers: [Whose calendars should also be checked]
- Time slot requirements: [e.g., next week, 45 minutes each, find 3 candidate slots]
### Optional Fields
- Interview format: [On-site / video?]
- Email tone: [Follow the style samples or be more formal?]
## Output Specification
One invitation draft per candidate: explain the role, list the available time slots, and ask them to reply to confirm; the time slots must come from real free time in the calendar.
## Constraints
- ✅ Time slots are checked against the calendar with no conflicts with existing events
- ✅ One draft per person, with personalized salutations and details
- ❌ Do not send any emails
- ❌ Do not create or modify events on the calendar; leave video meeting links for me to fill in myself🔧 Matching Criteria Feedback (tuning hard-requirement judgments)
Expected result: The Bot updates its criteria for "what counts as a match, what counts as unsuitable" based on your feedback and writes them into the skill, so the next list is noticeably more accurate—what should appear appears, and what shouldn't no longer does.
## Role
You are my recruiting sourcer and have just sent me a candidate list.
## Task
Based on my feedback, update the matching criteria for "what counts as meeting a hard requirement, what counts as unsuitable" and write it into your "job sourcing research" skill.
## Input Information
### Required Fields
- Mismatches: [List the ones that shouldn't have appeared and the missed ones, and state the correct judgment]
- Basis for judgment: [The corresponding hard-requirement clauses]
### Optional Fields
- New signals: [Background signals that should count toward matching going forward?]
## Output Specification
First restate the new criteria you've summarized; update the skill only after I confirm.
## Constraints
- ✅ Write the criteria into the skill so they take effect on the next sourcing run
- ✅ Only adjust the matching criteria; do not change the output format or approval boundaries
- ❌ Do not loosen hard requirements just to fill the headcount
- ❌ Do not contact any candidatesCheckpoint ✅
- Email and calendar show Connect apps as Installed (not Needs auth / Disconnected), and the ATS is reachable (via connector or browser login)
- In the first list, every match comes with evidence notes and people already in the ATS are excluded
- Throughout, no candidates were contacted and no records in the ATS were modified
- The sourcing workflow is saved as a skill and can be rerun by name with a different job description
- Interview invitations stay on draft cards, and calendar interview events are created only after your confirmation
Pitfall Reminders
❌ Matches have no evidence, just "this candidate is a great fit"
This means the hard requirements weren't written concretely. Evidence must be traceable back to the source—which hard requirement was hit, and which piece of information from which channel it's based on. Use the "Matching Criteria Feedback" magic Prompt to send it back a few rounds, and confirm that "evidence must be traceable back to the source" has been written into the skill.
❌ Candidates already in the ATS get sourced again
Two possibilities: ① The Bot can't read the full roster in the ATS—give it access to the view where the candidate pool lives, or state explicitly in the instructions "pull the ATS roster first, then source"; ② You later manually added people to the ATS—have the Bot re-pull the roster before every sourcing run instead of reusing the stale list from last time.
❌ The Bot goes to channels it shouldn't to scrape candidate profiles
The official docs are clear on this role: use "approved sourcing tools" and "respect the terms of each source." Write the list of allowed channels into the Bot description; to see candidates' public activity, use the connection-free X public content search (read-only), and don't let the Bot crawl platforms that require login or prohibit scraping by their terms.
❌ Wanting the Bot to contact candidates directly—not possible, don't even try
The official docs say it plainly: external outreach stays behind approvals; emails sent by routines are always saved as drafts waiting for you to hit Send, with no unattended-sending option. First contact also has hard rules: one email per person with one-click unsubscribe; if anyone unsubscribes or marks the email as spam, the Bot automatically stops sending to that address and tells you who it skipped. This isn't a flaw—it's the official design for the recruiting use case: your step-by-step approval process is precisely how you safeguard the candidate experience.
❌ Having the Bot add video meeting links to interview invitations
Official boundary of the Google Calendar connector: it cannot change Meet settings (waiting room, host controls, etc.), nor can it add Zoom links to invitations. The official workaround: fill in the link yourself in the event's location or description field, or create the event first yourself and let the Bot update the remaining fields.
Advanced Tips (Optional)
- Weekly sourcing routine: The official six-step method says to build routines only after retries and failure scenarios are defined. Schedule a weekly sourcing routine for a new role—keep it at the reviewable list and don't add approval-requiring actions (approval cards triggered by routines expire if not handled within 30 minutes). Routines run in the cloud, so they run even with your laptop closed. See 3.3 Routines.
- X public content as background research: The Bot can search and read public content on X without any connection (read-only, not logged in as you)—reading candidates' public posts and seeing who is publicly discussing the field are both ready-made background material; private content like DMs and bookmarks is what requires the X plugin. See 2.5 Connect Apps.
- Archive sourcing lists as Pages: Have the Bot organize each role's sourcing list into a Pages page (officially positioned as readable pages like reports and dashboards) and store it in the Bot's Library—for showing hiring managers progress, or looking back at who was sourced for a role, it's faster than digging through chat history. See 2.2 Files and Reviewable Results.
- Switch the information sources and you have a new scenario: using the same six steps, swap the ATS, sourcing tools, email, and calendar for a CRM, product usage, support tickets, and billing, and you get a ranked list of risk and upsell signals for A5 customer health.
Lesson summary
- ✅ One "Talent Scout" Bot + four types of data sources (ATS, sourcing tools, outreach mailbox, calendar, plus job descriptions and writing style samples), producing a reviewable AI recruiting sourcing list with evidence
- ✅ Start with the official starter prompt: find candidates who meet the hard requirements, exclude people already in the ATS, explain the evidence, draft personalized outreach—without contacting anyone
- ✅ Trace evidence back to its sources, refine until it's reviewable, then save it as a skill and test with a different job description
- ✅ Interview scheduling connects to the calendar: it can check free time slots, draft invitations, and create events, but invitations stay as drafts, calendar changes require your approval, and video meeting links are filled in by you
| Key item | Value in this lesson |
|---|---|
| Bot role | Talent scout—sourcing, candidate research, outreach drafts |
| Connected apps | ATS (connector or browser login), approved sourcing tools, Gmail/Outlook, Google Calendar |
| Matching criteria | Must-have hard requirements from the job description + an evidence explanation for each match |
| skill | "Job sourcing research" (name it yourself, invoke it by name, shared across all Bots) |
| routine | Weekly sourcing (optional, advanced), stops at a reviewable list |
| Approval line | Outreach produces a draft card first; routine emails are always saved as drafts awaiting your Send; privacy and source terms are written into the Bot description |
Next lesson preview
Next lesson A5 customer health: still the same six steps, swapping the information sources for a CRM, product usage, support tickets, and billing, letting the Bot watch your customer portfolio for risk and upsell signals on your behalf. You will learn:
- Combine recent usage, support escalations, renewal dates, and key-contact activity into one ranked watch list
- Each account comes with evidence, why it matters, and a suggested next step—no contacting customers, no CRM changes
- Define risk thresholds in the Bot description so weekly results stay consistent