Automate Interview Scheduling Communications

Workflow area: Automation · Tier: Core · Hands-on time: ~40 minutes

Interview comms are the perfect automation candidate: high volume, rigid structure, and painful when dropped — a candidate who never got the video link, a panelist who got no prep brief, are self-inflicted process wounds that cost real candidates. The design mirrors pb-36: the machine watches the sheet and drafts every message the moment a row changes; a human approves sends. Candidates are judging your company through these emails, which is exactly why they should be consistently excellent and human-checked.

When to use — and when not to

• Use it when you run more than ~5 interviews a weekbelow that, templates in your email client (pb-11 style) are enough.
• Use it when comms drop during busy weeksthe flow's whole point is that the day-before reminder goes out during exactly the weeks you'd forget it.
• Use it alongside a scheduling tool, not instead of oneCalendly or your ATS picks the slot; this flow handles the wrapper comms those tools do badly (prep materials, panel briefs, warm tone).
• Skip it if your ATS already does this wellGreenhouse/Lever tiers with full comms automation make this redundant; this playbook is for teams running recruiting from sheets and inboxes.
• Skip candidate evaluation content entirelythis flow schedules; it never scores. Keep feedback and assessments out of the sheet it reads (see the privacy note).

The recipe

1. 1.

   Build the tracking sheet — the single source of truth. One row per interview in Google Sheets: candidate_name, candidate_email, role, stage (e.g. screen/panel/final), interview_datetime, duration, format (video/in-person + link or address), panelists (names), panelist_emails, status, comms_log. The status column drives everything: scheduled → confirmed_sent → reminded → done, plus rescheduled and cancelled.

2. 2.

   Draft the message set once, in ChatGPT, before wiring. Five templates, one pass:

   Copy

   Draft our interview communication set. Company: [NAME, ONE-LINE
   DESCRIPTION]. Tone: warm, organized, respectful of everyone's
   time — we want candidates to feel expected, not processed.
   
   Use {{placeholders}} for: candidate_name, role, stage,
   interview_datetime (with timezone), duration, format_details,
   panelist_names, recruiter_name.
   
   1. CANDIDATE CONFIRMATION: what to expect at this stage, who
      they'll meet (names + roles), format and joining details, what
      to prepare (nothing surprising), who to contact for changes.
      Under 150 words.
   2. PANEL INVITE NOTE: goes with the calendar invite — role,
      stage, what this round assesses, link to the question bank
      section they own [SEE PB-30], and the rule: score
      independently before the debrief. Under 120 words.
   3. CANDIDATE DAY-BEFORE REMINDER: brief, warm, joining details
      again, "we're looking forward to meeting you". Under 80 words.
   4. RESCHEDULE (CANDIDATE-INITIATED): gracious, zero guilt, next
      steps to rebook. Under 80 words.
   5. RESCHEDULE (OUR SIDE): apologetic without groveling, concrete
      new options, acknowledges their time. Under 90 words.
   
   Rules: no jargon ("loop", "onsite" unexplained), no false
   urgency, nothing that promises an outcome or timeline for a
   decision unless placeholdered as {{decision_timeline}}.
   

   Edit until they sound like your best coordinator on a good day. These frozen templates go into the Zap — freezing them (rather than generating fresh per candidate) keeps quality consistent and review fast; the AI step personalizes only the variable details.

3. 3.

   Wire the Zaps — two small ones beat one giant one.

   Zap A — new/changed interview: Trigger: Google Sheets New or Updated Row where status = scheduled. Action: ChatGPT step — "Fill these two templates with this row's details; flag CONFLICT if any field is missing or the datetime is in the past" (paste templates 1–2). Action: Gmail — Create Draft to candidate; second draft to panelists. Action: update status = confirmed_sent, append to comms_log.

   Zap B — day-before reminder: Trigger: Schedule, daily 8 a.m. Action: Sheets lookup — rows where interview_datetime is tomorrow and status = confirmed_sent. Action: fill template 3 → Create Draft. Action: status = reminded.

   Reschedules stay semi-manual by design: change the row, set status = scheduled again, and Zap A re-drafts — you pick reschedule template 4 or 5 in the draft pass.

4. 4.

   Test with a fake candidate. Add a row for "Test Candidate" interviewing tomorrow with you and a colleague as panel. Run both Zaps. Verify: names and times correct including timezone, joining link present, panel note references the right stage, drafts created (not sent), the status column advanced, and a row with a missing field produced a CONFLICT flag rather than a confident blank. A good confirmation draft reads specific, not templated — check candidate name against email address match especially; mismatched merge fields are this flow's classic failure.

5. 5.

   Run the daily approval pass. Ten minutes each morning: open the drafts, check each against the sheet and calendar (did the panelist accept? did the candidate already email a change?), personalize one line where you know something ("Saw your question about the team — Priya will cover that"), send, done. The pass is also your error net: any draft that looks wrong means a sheet row is wrong — fix the row, not just the draft.

Make it yours

• ATS-lite upgrade: add columns for offer_stage comms and generate templates 6–7 (offer call follow-up, gentle check-in) — same flow, extended pipeline.
• Candidate experience touches: a template variant per stage that includes stage-appropriate extras — parking instructions for onsites, panel LinkedIn profiles for finals — one format column value switch.
• Panelist prep enforcement: add a Zap that checks, the morning of the interview, whether each panelist accepted the calendar invite, and drafts a nudge for anyone pending — the silent no-show killer.

Quality check

• Every send passed a human eyeboth Zaps end at Create Draft; verified during the fake-candidate test.
• Names, times, and links match the sheet exactlywith timezones explicit; the classic mismatch was tested for.
• Missing data produces flags, not blanksthe CONFLICT rule fired on the incomplete test row.
• The status column is the single source of truthno message goes out without the sheet reflecting it in comms_log.
• Nothing promises outcomestemplates commit to logistics only, or use the {{decision_timeline}} placeholder a human fills deliberately.