Candidates take AI screening interviews after hours

What the data shows

Tengai, a Swedish recruitment technology provider, analysed 20,291 screening interviews that candidates completed with its AI interviewer between January 2025 and September 2026. 44% of them started outside office hours, which the company defines as before 8am or from 5pm on weekdays, plus the whole weekend. About 15% took place at weekends, and Sundays (8.9%) were busier than Saturdays (6.4%). Roughly one interview in four started after 6pm.

Late-night sessions were rare. Only about 7% began between 10pm and 2am, and weekday daytime was still the most common slot, with a peak between 11am and 3pm. Candidates did not abandon the working day. They used the evening and the weekend when the daytime did not suit them.

What it does not prove

This is one vendor's data from one country, published by that vendor. Onrec, which reported the analysis, says plainly that it is not a representative measure of job interviews in Sweden or the UK. It also says the figures do not show whether AI screening produces better hires. Nothing here tells you that candidates prefer talking to software over talking to a recruiter.

The figure also counts only people who were offered an on-demand interview. Candidates who dropped out before the screening step, or who never applied because the process looked awkward, are not in it. Treat the 44% as a sign of when people can make time, not as proof that automation is what they want.

Why timing matters to an agency

For an agency, the first screen is often the narrowest point in the process. A phone screen needs the candidate and the recruiter free at the same moment, and that moment usually falls inside the recruiter's working day. Tengai's own press release makes this point: the traditional phone screen asks both sides to be available at once, during work hours.

The people this hurts are often the ones clients want most. A candidate who already has a job, works shifts, or collects children from school may not be able to take a call at 2pm. If your first step only runs in daytime, some of them drop out quietly, and you never learn why. Onrec draws the same conclusion: fixed daytime appointments are hard for people in work or with caring duties.

One change to make

You do not need an AI interviewer to act on this. The practical change is to let the candidate pick the time of the first screen, inside a window that includes some evening and weekend slots. That can be a structured video question set the candidate records when it suits them, a self-booking calendar with a few early or late slots, or an AI screen where the client has agreed to one.

Keep three things in place whichever route you take:

• the same questions for every candidate, so an evening answer is judged on the same basis as a morning one;
• a recruiter, not the tool, deciding who moves forward;
• a human alternative for anyone who does not want an automated step or needs an adjustment.

Then measure it. Compare completion rates for the first screen before and after the change, and note what share of screens now happen outside your office hours. That gives you your own figure rather than a vendor's, and it is the number a client will actually trust.

The takeaway

The Swedish data does not show that AI makes hiring better. It does show that, given the choice, many candidates fit the first conversation around their own day. An agency that still runs its first screen only from nine to five is asking working candidates to take time off before they have even spoken to anyone. Moving that one step out of the office calendar is cheap, and you can test it on your next role.