Panel interviews and interview pools Availability without calendar access Integration with your ATS

Custom interview scheduling software development

Scheduling a single interview is a calendar problem. A panel interview with three interviewers, a candidate who gets to choose, and interviewers who also have their own work to do is a scheduling problem with hard constraints. Appfront builds scheduling software that solves it: valid proposals instead of open gaps, a rescheduling route when someone cancels, and control over how interviews are spread across your interviewers.

Why scheduling one interview is easy and a panel is not

Interview scheduling software determines which interview slots are actually possible, offers them to the candidate and puts the confirmed appointment in all the calendars involved. If you are looking for the full recruitment process, with candidate tracking, vacancy management, matching and job boards, see building recruitment software; scheduling is one part of that, and at low volumes it is also the right place to start. If you run panels and schedule dozens of interviews a week, scheduling becomes a specialism that justifies a solution of its own.

Scheduling a one-hour interview between two people is trivial: take the overlap of two calendars and pick the first gap. A panel interview with three interviewers is something else. You need a continuous block of sixty or ninety minutes in which all three are free, within everyone's working hours and time zone, with a meeting room or video link, buffer time between rounds, and the restriction that one of them can do at most two interviews a week because they also have their own job. Add a loop of three rounds on a single day and the number of combinations grows quickly.

This is the same class of problem as timetables and shift rosters: combinatorial, with hard constraints and soft preferences. There is no method that always finds the best solution quickly; in practice you put a constraint solver on it, such as the open source CP-SAT model from Google OR-Tools. That is the difference from a calendar link: a calendar link shows gaps, whereas a scheduling system only shows combinations that have been proven to work.

Hard constraints

All panel members free in the same block, within their working hours, with a room available. A proposal that breaks any of these only shows up as an error once it reaches the candidate.

Soft preferences

Preferably not three interviews in one day for the same assessor, preferably a panel with mixed experience. They count towards the score, but block nothing.

Where interview scheduling gets stuck in practice

A schedule that is in place rarely stays that way for long. An interviewer falls ill, a client meeting moves. With a one-to-one interview that is a minor reshuffle. With a panel, losing that one member brings down the whole block: the other two see their blocked time freed up, and the candidate has taken half a day off for it.

The question that follows is organisational, not technical. Do you find a replacement from the same pool, reschedule the whole block, or carry on with two assessors? A replacement is only possible if it is already defined which competencies that round assesses and who else can cover them; otherwise you quietly change your selection process. Rescheduling the block means solving the scheduling problem again, with less room to manoeuvre. Carrying on with two can be defensible, provided someone is authorised to approve it and records the deviation. Software that does not support this pushes the decision back to the recruiter, who is busy with ten other things at that moment. That is the difference from a calendar link: not in the scheduling, but in the rescheduling.

Interviewers are the scarce resource, not the calendar

Without guidance, every scheduling algorithm, and every coordinator in a hurry, picks the people with the most free space in their calendars. These are invariably the same handful of colleagues. They are not rewarded for it, and it comes at the expense of their own work. After a few months they decline more often or drop out of the pool, just when you are scaling up. A scheduling system should handle capacity explicitly: a weekly maximum, visibility of the actual distribution and a pool per round type. Standard packages partly cover this; Greenhouse works with interviewer groups, working hours, buffer time and per-interviewer limits.

Lead time is part of your offer

In the first quarter of 2026, according to the UWV tightness indicator, the labour market was tight or very tight in 87 of 93 occupational groups. If you wait three days to reply in a market like that, the candidate is already sitting at another company's table. The time between "we'd like to speak to you" and a confirmed appointment is therefore not an administrative detail.

What a scheduling system needs to do

Reliable availability and a solver are the foundation. Four components determine whether the system holds up in everyday use.

Candidates choose for themselves, with a hold

The candidate sees several realistic slots, including the interviewers involved. While the choice is open, the system holds those slots and then releases them automatically.

Rescheduling as a core feature

When someone cancels, the system calculates the impact on the whole block and offers the options: a replacement from the pool, rescheduling the block or continuing with fewer assessors.

Pools and load balancing

A pool of qualified interviewers per round type, with a weekly maximum. New assessors shadow interviews until they are cleared to be scheduled independently.

Feedback to the ATS and to people

The confirmed appointment lands with the candidate in the ATS, interviewers receive their invitation with attachments and video link, and every change remains traceable.

Retrieving availability without looking into calendars

To schedule, you need one thing from an interviewer: when they are busy. Not with whom, about what or where. That distinction can be enforced technically, and it is the first question your works council or privacy officer will ask. Google Calendar has the freeBusy endpoint for this: only the start and end times of busy blocks, never titles or invitees, for up to fifty calendars per request. There are separate OAuth scopes for it, so you do not need to ask for read access to the full calendar.

In Microsoft 365 the same works through getSchedule in Microsoft Graph, with Calendars.ReadBasic as the least privileged permission. The response is an availabilityView: status codes per time slot, where 0 is free, 1 tentative, 2 busy, 3 out of office and 4 working elsewhere. You get working hours and time zones from findMeetingTimes or the mailboxSettings. Watch out for the pitfall that an app-only integration can by default access all mailboxes in the tenant; you limit that scope with RBAC for Applications, the successor to application access policies.

We send invitations via iCalendar (RFC 5545) and iTIP (RFC 5546), so a rescheduled appointment updates in Outlook or Google Calendar instead of ending up as a duplicate.

  • Free/busy information only, no appointment details
  • Google freeBusy with the calendar.freebusy scope
  • Microsoft getSchedule with Calendars.ReadBasic
  • App access limited via RBAC for Applications
  • Audit log of every scheduling action
Google Calendar API (freeBusy) Microsoft Graph (getSchedule) iCalendar RFC 5545 iTIP RFC 5546 CalDAV RFC 6638 OAuth 2.0 OR-Tools CP-SAT ATS APIs

Even without appointment content, you are processing applicants' personal data. The Dutch Data Protection Authority (Autoriteit Persoonsgegevens) expects that data to be deleted within four weeks of the procedure at the latest, or after one year at most with consent. A system that does not enforce this quietly turns into an archive of rejected candidates.

Not sure about a large project yet?

Test your idea first: a working prototype in 1 day

With OneDayBuild, we make your idea tangible in a single day for €1,150, so you know whether further development is worth the investment. Decide to go ahead with the full build? Then we deduct the cost in full.

View OneDayBuild →

When you should not build this

For a good share of readers, a custom build is the wrong choice here. If you hold around ten interviews a month, almost always one-to-one and with the same hiring managers, a scheduling link such as Calendly, Microsoft Bookings or Cal.com will take you a long way.

If you already have a mature ATS, that is where the investigation starts. Interviewer groups, working hours, buffer time, limits and self-scheduling are built into the better packages, and a fair share of the organisations considering a custom build have never configured them. If you run panels at volume but without unusual rules, there are scheduling layers designed for exactly that which connect to the well-known ATSs, such as GoodTime and ModernLoop. Building the same thing yourself only adds maintenance.

A custom build only starts to make sense when your scheduling rules fall outside what those products can handle: interviews that may only be conducted by certified assessors, professionals who work to a roster and can therefore only attend outside their shifts, or a panel that must include a member from outside the department. An ATS without a usable scheduling API is also a reason, and once coordination hours have become the biggest cost item, the calculation changes.

If you build it yourself, factor in the burden that comes with it. Google's calendar scopes are classed as sensitive and require verification of your application; in Microsoft environments you need administrator consent per tenant. And responsibility for retention periods and the record of processing activities lies with you.

One more boundary, concerning AI. As long as the system only calculates time slots, it remains a scheduling tool. If you add assessment, such as ranking candidates or having interviews analysed, you fall into the recruitment and selection category of Annex III of the AI Act. That is a separate matter, which we cover in AI for HR and recruitment.

Consider custom development when

  • Panel composition rules that your package does not support
  • Your ATS offers no usable scheduling API
  • Coordination hours have become the biggest cost item
  • You are stuck on your vendor's roadmap

Related services

Two topics almost always come up alongside this one.

Custom recruitment software

If you are looking for the full recruitment process, with candidate tracking, vacancy management, matching and job boards, see building recruitment software. Scheduling is just one module in that larger whole.

AI for HR and recruitment

As soon as assessment comes into play, from ranking candidates to summarising interview notes, different rules apply. What is sensible and permitted there is covered in AI for HR and recruitment.

Frequently asked questions about interview scheduling software

At low volumes and one-to-one interviews, that is exactly what you should do: a scheduling link or your ATS's scheduling feature covers it fully. A custom build only becomes defensible for panels, high volume or rules that no package supports.

It doesn't have to. Scheduling only needs free/busy information: the start and end times of busy blocks. Google provides this through the freeBusy endpoint with its own scope, and Microsoft 365 through getSchedule with Calendars.ReadBasic. Titles, attendees and locations never come into it.

With a panel, a single cancellation brings down the whole slot, so handling it is a core part of the design. The system offers the options: a replacement from the same pool, moving the slot, or going ahead with fewer assessors. Who may choose which option is defined per round type.

Yes, and this is often the strongest reason to build. Without control, interviews end up with the colleagues who have the most open calendars, and they eventually drop out. The system therefore works with a pool per round type, a weekly maximum and visibility of the distribution.

With Google Workspace and Microsoft 365 calendars, with Teams, Google Meet or Zoom for interview links, and with your ATS for candidate data and status feedback. Invitations run via iCalendar and iTIP, so rescheduled appointments are updated rather than left behind as duplicates.

We agree up front that you own the source code, and we deliver it in a repository you control, with documentation of the architecture, integrations and scheduling rules. We record those rules explicitly rather than burying them in code: they change more often than the technology underneath.

A system that only calculates time slots is a scheduling tool, not an AI system for recruitment and selection. If you add assessment, such as ranking or pre-selecting candidates, it falls under Annex III of the AI Act; the application date for that has been pushed back to December 2027 through the Digital Omnibus.

Work through your scheduling challenge first

Tell us how many interviews you schedule per week, how your panels are put together and which rules apply. We first check whether your ATS or an existing product can already do it. If custom development turns out to be the route, we map out constraints, integrations and rescheduling policy before a line of code is written.

Edit content