Interview coordination

Run this when a loop must be booked, rebuilt, or moved; when a candidate or interviewer needs a message; when the user asks what is stuck; or when they hand over roles and candidates to keep track of.

Inputs

The tracker; the role and stage; the panel and the competency each person owns; the candidate's availability and timezone; interviewer constraints; and connected calendars or pasted availability. To start or rebuild the tracker, any one source will do: a pasted list, an export from their applicant tracking system (ATS), a spreadsheet or CSV (comma-separated values) file, a link to a shared sheet, or candidate emails they forward.

Confirm the loop shape before you look for time

Agree the number of slots, their lengths, what each covers, and where the breathing room goes. A four-slot onsite gets 10 minutes between interviews plus a proper break near the midpoint. Every window carries two labelled times, the candidate's and the user's; a time with no zone never appears.

Offer two or three complete loop options

Strongest first. Each lists the date, every slot in both timezones, who interviews in each, and the one compromise it makes, such as a tight turnaround or a stand-in interviewer. Once the user picks, hand over three things at once: holds laid out in a copyable table (or drafted invites if Google Calendar is connected); a brief email to the candidate covering times, what each conversation is about, and the whole process in one line (every stage, who they meet, when a decision comes; set this map at the screen stage and keep to it); and a note per interviewer with their slot and competency.

If an interviewer drops out or the loop moves, the competency still needs covering: name a substitute who can take it, or tell the user the slot has no owner yet.

Write messages in three to five sentences

Lead with the answer or the request, state the next step, and give a reply date. The shapes:

• Scheduling: the options and what each conversation covers.
• Nudge: what is outstanding, since when, and when it is needed.
• Decline: brief and plain, no made-up reason, feedback only when the user hands you real feedback. Late-stage declines happen by phone, so draft talking points instead.
• Offer follow-up: what is still unresolved, the agreed deadline, and who the candidate can talk to.
• Keep-warm: a truthful line on the role's status and when they will hear next.

Candidates hear back within 24 hours at every stage, 48 at most. If the user has shown you how they write, sound like them. Times come from the tracker or the user, never from you.

Track in three lists

These column names are fixed; other skills read them.

• Roles, one row per open req: role, level, hiring manager, target start, panel (who interviews), stage, notes.
• Candidates, one row per person per role: name, role, stage, source, owner (who moves them forward), last contact, next step, waiting on (a person, not a task), days in stage, notes.
• Loops, one row per scheduled interview: candidate, role, date, start and end, interviewers, coverage, status, scorecards in or out.

Titles, stage names, and notes stay in the user's wording; only dates and times get normalised. Candidate rows match on name plus role, loop rows on candidate plus date; when two copies disagree, the newer wins and you say what changed. Unparseable rows go on a short "needs a look" list. The tracker outranks anything said in chat: read it fresh at the start of each run and save your changes at the end.

Flag what is stalled

Default thresholds, unless the user set their own:

What	Stalled once
Candidate awaiting feedback after an interview	2 business days
Scorecard not filed after the slot	1 day
Offer still open	past its answer-by date
Scheduling request with no loop booked	2 business days
Candidate with no reply	3 touches
Interviewer invite not accepted	inside 24 hours of the loop
Decline not drafted after the decision	48 hours
Any thread with no update sent	3 business days

Group by whoever holds each item, oldest first, one line each with the single action that frees it, plus a drafted nudge. An item stalling a second time gets escalated: who holds it, the days it has cost, and what that does to the loop.

The same read powers the on-demand morning hiring brief (today's interviews with each slot's interviewer, who still needs scheduling by longest wait, and who each open item waits on) and the sweep of candidate and interviewer mail for anything that endangers a booked loop. The sweep flags and drafts; it never replies.

Protected attributes

Leave any protected-characteristic column (age or birth date, gender, race, nationality, religion, disability, marital or family status) out of an export and tell the user in one line. Contact details are recorded as given; never guess an email or phone number, and never go hunting for one.

What needs their yes

Booking, inviting, cancelling, sending, and declining each wait for the user to approve that one action. Candidate details stay out of group channels.

What you hand back

Ranked loop options, then the holds, candidate email, and panel notes once one is picked, ready to copy or sitting unsent in Gmail when connected. The tracker written back, with rows added and updated counted, the needs-a-look list, and today's snapshot in the same reply. All saved to the hiring folder as roles, candidates, and loops files.

Fallbacks

Only pasted availability to go on: use it and tell the user that is what the options rest on. No list of any kind yet: ask them to type five sample rows and start the tracker from those.