Willo

Company overview

Willo is a Glasgow, UK-founded (2018) async video interview company that screens candidates by collecting recorded video, audio, multiple-choice and text responses on candidates' own time. Co-founded by Euan Cameron, it grew during the remote-hiring boom, opened a New York presence, and serves startups through enterprises globally. Its AI layer ('Willo Intelligence') transcribes responses, generates structured summaries, ranks candidates against an employer-defined 'Blueprint,' and runs 'Real Talk' detection of scripted or AI-generated answers. Willo positions itself as 'human-first,' stating its AI summarizes and informs rather than scores video, and that algorithmic auto-rejection is limited to factual multiple-choice questions.

Regulatory exposure

Willo's ranking, Blueprint matching, and multiple-choice 'Benchmark' auto-rejection make it an Automated Employment Decision Tool under NYC Local Law 144 and a likely high-risk system under EU AI Act Annex III (employment) when deployed in scope. Despite this exposure, no public NYC LL 144 bias audit, no independent fairness audit, no EU AI Act / FRIA deployer materials, and no AI-specific technical documentation (model cards, ISO 42001) could be found. Willo mitigates risk primarily by design narrative — no facial recognition, voice scoring, or biometric analysis, and human-in-the-loop framing — rather than by published, jurisdiction-specific compliance artifacts. Its strongest verifiable posture is information security (ISO 27001:2022) and explicit human-oversight defaults.

Path to a higher score

The highest-value move is commissioning and publicly posting an independent NYC LL 144-style bias audit (e.g., Warden AI, Holistic AI, BABL AI) for the ranking/Benchmark features, renewed annually. Willo should publish a formal AI/model card or technical pack documenting the Intelligence and Real Talk models, their inputs, known limitations (e.g., accent/ASR fairness it already acknowledges), and explainability methods, and pursue ISO 42001. Adding explicit deployer guidance — candidate notice templates, opt-out/alternative-process language, an EU AI Act Art. 27 FRIA template, and a downloadable DPA/sub-processor page — plus a public status/incident channel and security contact would convert its already-strong human-oversight and data-security story into citable, jurisdiction-specific compliance evidence.