4 Interviewers, 3 Calendars, 12¢: The Panel Booker monday
Your SaaS scheduler is lying to you. It says it books interview panels automatically, but it quietly assumes every interviewer lives on the same calendar system. The moment one partner is on Fastmail and another is on Outlook, the whole thing collapses back to email ping-pong. If you're a founder or ops lead trying to hire seniors and you're spending two days a week chasing four calendars, this is for you. I'll show you the exact n8n workflow that resolves four calendars across three systems in under three seconds and books a panel for twelve cents. I'm Lazar, I build these systems for hiring teams every week. Here's the part every scheduler vendor hides.
Let's start with the actual failure mode, because until you see it you'll keep buying tools that don't fix it. Picture a boutique consultancy, four partners, they hire eight to twelve seniors a year. Partner one lives in Google Workspace. Partner two is on Microsoft 365 because of a legacy client. Partner three uses Fastmail with CalDAV because she values privacy. Partner four floats between two Google accounts. A candidate clears the screen and now you need to book a forty-five minute panel with all four of them inside the next two weeks. The old process took two point three days on average and eighteen back-and-forth emails per candidate. Multiply that by twelve hires a year and you've burned an entire month of coordinator time on nothing but scheduling. Now here's what happens when you try to fix it with the obvious tools. monday Recruiter, Calendly Teams, every shiny AI scheduler out there, they all quietly require one thing. Everyone has to be on the same calendar ecosystem, or you have to force the outliers to connect their calendar through some third-party sync tool that breaks every three weeks when a token expires. The tool doesn't fail loudly. It just silently ignores partner three's real availability and books over her yoga class. She shows up angry, the candidate gets rescheduled, and now you're back to email. The problem isn't back-and-forth email. The problem is cross-system availability. Every scheduler that markets itself as an AI recruiter is solving the wrong problem because solving the real one is boring plumbing and they can't put it on a landing page. Okay so here's what we build instead. The stack is n8n self-hosted, which means zero per-seat cost. Three calendar nodes running in parallel. For Google interviewers, the native Google Calendar node hitting the free-busy endpoint. For Outlook, the Microsoft Graph node hitting the same free-busy query. For Fastmail or any CalDAV calendar, a generic HTTP request node pulling the ICS feed with basic auth. All three fire in parallel, they return busy intervals for the next fourteen days, and they dump into a merge node. Step two is the interval intersection. This is a forty-line Python function node. You take each interviewer's busy blocks, invert them into free blocks, then intersect all four sets to find windows where every single person is free. Then you slide a forty-five minute window across those intersections and you emit every valid start time. On four calendars over a two-week horizon this runs in under two hundred milliseconds. No AI needed for this part. It's just math, and math is cheaper and more reliable than an LLM for anything deterministic. Step three is where Claude Haiku earns its twelve cents. You feed it the list of valid slots plus a small ranking rubric. Avoid the thirty minutes after lunch because panel energy is dead. Avoid Friday after three because half the candidates mentally check out. Cluster the senior partner first in the sequence because their read carries the most weight. Prefer Tuesday through Thursday. Haiku returns the top three slots ranked, plus a two-sentence rationale you can log for audit. Total cost per call, about eight cents. Step four, the same Haiku call drafts the candidate email. Three slot options, timezone-aware, with a one-click confirmation link that writes back to the workflow. Step five, when the candidate picks a slot, n8n creates the calendar invite on all four interviewer calendars using each one's native protocol, writes the booking record into the ATS via API or a simple Postgres row if the ATS is thin, and sends a Telegram ping to the coordinator saying panel booked, here's the slot, here's the candidate one-pager. Total per-panel cost, twelve cents in Claude calls. Infrastructure cost, zero, because n8n is running on a home server that already exists. Compare that to Calendly Teams at sixteen dollars per seat per month times four seats plus monday Recruiter at whatever they're charging this quarter, and you're looking at a payback period measured in days. Let me show you the actual booking loop. On the screen right now, a candidate has just cleared the technical screen. The workflow triggers. You can see the three calendar nodes light up in parallel. Google returns in eight hundred milliseconds, Graph returns in a second and a half, CalDAV returns in two seconds because Fastmail's endpoint is slower. Merge node fires. Python function runs the intersection, finds seventeen valid forty-five minute windows across the next ten business days. Haiku ranks them, picks three, drafts the email. Total elapsed time from trigger to email sent, fourteen seconds. The candidate replies with slot two. Four calendar invites go out, ATS gets updated, coordinator gets the Telegram. Zero human touches. Now the honest caveats, because I don't sell you fairy tales. CalDAV endpoints can rate-limit you if you poll too aggressively, so cache free-busy for fifteen minutes. Microsoft Graph tokens expire and the refresh flow is annoying to set up the first time, budget two hours for that. And if one of your interviewers has a shared assistant-managed calendar, you need to make sure the assistant is actually keeping it current, because no automation fixes a stale source of truth. But once it's running, it just runs. My client has booked forty-one panels through this workflow in the last six months. Zero double-bookings. Average time from clear-to-screen to panel-on-calendar dropped from two point three days to fourteen minutes.
Want more like this?
I publish practical AI automation, GenAI engineering, and faceless content workflows on YouTube every week.
Subscribe to bizflowai.io on YouTube — never miss a new tutorial.
Planning an AI automation project or need a second opinion on your architecture?
Connect with me on LinkedIn — Lazar Milicevic, GenAI Engineer & bizflowai.io Founder.
Visit bizflowai.io for our services, case studies, and AI consulting.