People
Interview scheduling, including the reschedules
Four calendars, two time zones, a candidate who can only do early mornings, and a panellist who declines twice. This is the job.
What the worker does
An AI worker in Polaris handles interview scheduling by reading panellist availability in Google Calendar, proposing slots that fit every constraint you stated, and drafting the invite and the candidate email for a person to send. When someone declines, it proposes the next set of options in the same task thread. The worker never emails a candidate directly.
- Connections
- Google Calendar · Gmail
- Handles
- Proposals, conflicts and reschedules
- Sends to candidates
- Never. A human does
Scheduling is not hard. It is just relentless.
A single onsite loop can take fifteen messages. Someone reads four calendars, picks three options, sends them, waits, loses one option to a meeting booked in the meantime, and starts again. Multiply by six candidates and a week is gone.
The reason it stays manual is that the constraints are unwritten: this panellist should not do two interviews back to back, that one is genuinely unavailable on Fridays regardless of what the calendar shows, and candidates currently employed elsewhere should be offered early or late slots so they are not explaining an absence.
Write those constraints down once, in the worker's skill file, and the relentless part goes away while the judgment stays with you.
Constraints worth writing into the brief
Each of these is an unwritten rule in most teams, and each is the reason a scheduling tool gets abandoned.
Buffers, not just gaps
Fifteen minutes either side of an interview for notes. A slot that touches another meeting is not a valid slot.
Panel load
No panellist does more than two interviews in a day, and never two in a row.
Candidate-friendly hours
Offer early morning or after 17:00 to candidates in another job unless they have said otherwise.
Time zones stated explicitly
Every proposed slot appears in both the candidate's local time and the panel's, written out, not implied by a calendar invite.
A hard latest date
If no slot is agreed within your target window, escalate to the hiring manager instead of continuing to propose.
What happens at each turn
| Event | Worker | Human |
|---|---|---|
| Interview requested | Reads calendars, proposes three slots with time zones | Picks or amends |
| Candidate replies with a preference | Checks it still works, drafts the invite | Sends it |
| A panellist declines | Proposes replacement slots in the same task thread | Confirms |
| Candidate goes quiet | Flags on the agreed day. No chasing mail sent | Decides whether and how to follow up |
| Interview happens | Books the debrief slot if that is in the brief | Runs the interview and the decision |
What you are actually buying
Not a scheduling widget. The hours around one.
$0
For the workspace, per person
No seats, unlimited humans
~$2
Per human-hour delivered
Only when a job actually delivers
0
Emails sent to candidates by a machine
Every candidate message is sent by a person
Questions people ask
+Does it need access to every panellist's calendar?
It needs the Google Calendar connection authorized for the org and visibility of the calendars you want it to read. Connections are authorized once and stored server-side, so nobody is pasting credentials into a task.
+What about candidates who are not in Google Calendar?
They never are. The worker reads your side and proposes times, and the exchange with the candidate happens over drafted email that a person sends. That is the same way a human coordinator does it.
+Can it handle a full onsite loop with five panellists?
Yes, and that is where the constraint list earns its keep. Give it the buffer rules, the load limits and the order the panel should run in, and check its first proposed loop closely before trusting the next ten.
+Where do reschedules live?
In the same task thread as the original request, as further comments. The whole history of a loop sits in one place instead of across a mailbox, a calendar and a chat channel.
Related
HR work an AI worker can prepare, and where it must stop
Scheduling, onboarding logistics, policy drafts and review-cycle admin, prepared by a worker. Every decision about a person stays with a person.
Application summaries, not candidate scores
A worker reads every application in the same shape against the criteria you published. It does not rank anyone, and it does not reject anyone.
Onboarding that is finished before the first Monday
Nobody's first day should start with an apology about accounts. A worker runs the pre-start checklist and reports what is not done while there is still time to fix it.
Hire an AI executive assistant
Every commitment made in a meeting becomes a task with a name and a date on it before the next meeting starts.
Hire an AI recruiter
Every candidate is scored against the same written rubric, and the rubric is a file your hiring manager wrote and can change.
Connect Google Calendar to Polaris
Scheduling is the one thing an assistant is asked for first, and it is the one connection still waiting on Google sign-in.
Connect Gmail to Polaris
The honest version: Gmail is in the catalog, the Google sign-in flow has not shipped, and the product says pending rather than pretending.
Every commitment from the meeting, owned and dated
The decisions were good. The problem is the eleven commitments that existed only in a document nobody reopened.