The hiring stack was built for a world that no longer exists.

Applying for a job used to be hard. You had to find a printed ad, type a cover letter, walk into a post office. The filter was the effort of applying. Every applicant who reached the top of your pile was already self-selected.

That world ended around 2015. Today a candidate with one afternoon and ChatGPT can submit fifty tailored applications before lunch. Three hundred applications per requisition is normal. A thousand is common. Ten thousand happens. The human being who reviews them has the same forty-five-minute window they had in 1998.

Recruiting teams respond with the only tool they have — heuristics. Keyword filters. University filters. Tenure filters. You end up rejecting good candidates because their resume isn't search-optimised, and interviewing mediocre ones because their keyword density is high. Everyone knows this. No one has a better answer.

The better answer is to talk to everyone.

That sounds absurd until you do the math. A thirty-minute structured interview is the most predictive thing you can do before hiring someone. The reason recruiters don't interview every applicant isn't principle — it's capacity. There are not enough people-hours in a week to screen a thousand candidates properly.

So we built the people-hours. Buster — our AI interviewer — runs a proctored, multi-turn, rubric-scored interview with every applicant. Not a keyword filter. Not a sentiment score. A real conversation, adapted to the job description you wrote, proctored well enough to spot tab-switching, scored against criteria you can inspect and change.

The output isn't a yes/no. It's a full transcript, a per-competency breakdown, three direct quotes, and a weighted score you can trust because the rubric was built from your own JD. That output is what lands in your kanban the next morning.

Why not just use ChatGPT?

Because hiring is not a one-shot prompt. It's a workflow with seven stages, five roles, three languages, an SSO policy, a bounce-handling suppression list, and a legal obligation to reject people kindly. DawnStep is the scaffolding that makes the AI interview actually useful — the triage, the kanban, the debrief panel, the audit trail, the SES-backed email, the Teams transcripts, the per-tenant SSO configuration.

The interview is the unlock. The product is what makes it operable.

What we believe

• Every candidate deserves a real conversation. Not a keyword filter. A conversation adapted to the job you're actually hiring for.
• The recruiter stays in charge. Buster triages. People hire. Every automation is logged, reversible, and overridable.
• The rubric is visible, not a black box. You see it, you edit it, you re-score against it. No hidden models making decisions you can't explain to a candidate.
• AI is how, not what. We don't sell AI. We sell a hiring loop that happens to run on it.

Who built this

DawnStep is built by AT Dawn Technologies, a small team that got tired of seeing good engineers rejected by keyword filters and mediocre ones advance because they knew how to game them. We started with the hiring loop we wanted for ourselves. It turns out a lot of teams want the same thing.