Most AI hiring tools are built to decide for you, to score, rank, and filter people automatically. We built Virvell to refuse. A hiring decision changes the course of someone's life, and that is not a choice to hand to a black box.
AI collects the data. You decide.
The commitments the product already keeps.
Virvell does not score, rank, or recommend candidates, and it never rejects anyone automatically. Not as a default, not as a setting, not as a feature anyone can pay to unlock. If a customer asks us to switch it on, the answer is no. We would rather lose the deal than the principle.
We hold that line for a simple reason. "The AI recommended it" is not a defense, to a candidate or to a regulator. As regulation tightens, with New York and Ontario among the first, the tools that score and filter people automatically will be the ones called to account. We give you more evidence and less algorithmic opinion.
When the calls are complete, we compare them. Does the resume match the pre-screen? Does the pre-screen match the reference? We surface the discrepancies worth a second look, a title that has grown, a scope that has shifted, and we trace each one back to where it came from. We do not tell you what it means. We show you what was said, where it came from, and where it does not line up. The decision stays yours.
High completion rates come down to convenience. A candidate can call back at nine in the evening, on their own schedule, instead of leaving a shift to make a one o'clock window. The person calling back at night is the one the old process lost. Convenience is not a feature here. It is the fairness argument.
We are equally clear about the limits. A voice-first process is not automatically inclusive, and it can disadvantage people it was never meant to. Naming that is the first step to fixing it, and we come back to it below.
At the start of every call, the AI identifies itself as AI, states that the conversation is recorded, and asks for consent before continuing. Anyone may decline. The call ends there, and they are directed to the hiring team. We keep that promise even when it costs us a completed call. Our AI Acceptable Use Policy is public and written in plain language, and we point buyers, candidates, and their counsel to it directly. Transparency is not a feature we added. It is where we began.
Read our AI Acceptable Use PolicyWe are early, and we would rather name what we are still building than pretend the work is finished.
The question that matters most to us is also the hardest: could a voice-first process disadvantage anyone, neurodivergent candidates in particular, and how do we ensure it does not? If the aim is to find the best people, excluding someone for how they speak rather than what they say is a failure by any measure. When inclusion and speed pull against each other, we slow down.
This fall we are engaging an independent AI ethics and readiness advisor to examine our foundations, our approach to bias, and our accessibility, with formal third-party certification as the longer-term goal. The intent is to do this with outside accountability and current research behind us, not on our own assurance.
Completion data shows us how candidates move through the process. It does not tell us how they experienced it. So we are building a direct channel for that feedback, designed around a single rule: what a candidate tells us about the experience stays separate from their evaluation and is never used to assess them. No buyer will ever see that separation. We are building it anyway, because it is right.
Building AI for hiring carries real responsibility, and the standards for doing it well are still being written. A tool that refuses to score people, when nearly every competitor does, is harder to build. We take that path on purpose.
If you want to hold us to any of this, the policy is here, and so are we.