impress.ai

Company overview

impress.ai, operated by Ideatory Pte. Ltd. (80 Robinson Road, #08-01, Singapore 068898), is a recruitment automation vendor founded in 2017 with regional presence in the US, Australia, India and Indonesia. Its platform — now fronted by an agentic product line called Savos — combines conversational AI assistants, resume parsing and relevance scoring, structured competency-based pre-screening, assessment orchestration, asynchronous video interviewing and scheduling, and integrates with major ATS platforms including Workday, SAP SuccessFactors, Greenhouse and Taleo. Publicly named customers include Singtel, Apex Group and Singapore public-sector agencies, and the company launched a lighter mid-market product, impress.ignite, in 2024. Pricing is not published anywhere on the site.

Regulatory exposure

The product automatically scores, ranks and shortlists job applicants, placing it squarely inside EU AI Act Annex III high-risk employment use and inside the NYC Local Law 144 AEDT definition; the asynchronous video interview surface adds Illinois AIVI exposure, and consequential-decision screening adds Colorado AI Act exposure. As a provider, impress.ai would carry Article 11 technical documentation, Article 13 transparency, Article 14 human oversight, Article 15 accuracy and robustness, and Article 17 quality-management duties for EU deployments. Nothing published addresses any of this: there is no EU AI Act positioning, no bias audit, no FRIA or deployer guidance, and no AEDT or LL 144 notice material. The company's own blog concedes it "provides both GDPR compliant and non-GDPR compliant versions," and its product pages market a "Bias-Free Screening Process" and unbiased candidate evaluation without publishing any adverse-impact testing to substantiate the claim. Its compliance posture rests almost entirely on information-security certification (ISO 27001/27017/27018, ISO 9001, SOC 2 Type II) rather than AI-specific governance — a meaningful mismatch for a tool whose core function is algorithmic candidate selection.

Path to a higher score

The highest-value single move is commissioning an independent bias audit from a recognised auditor (BABL AI, Holistic AI, DCI Consulting, Warden AI or ORCAA) covering EEOC race, ethnicity and sex plus the intersectional categories LL 144 requires, and posting the summary as a freely downloadable document — this would both close the largest gap and substantiate the fairness marketing the company already runs. Beyond that: publish a system or model card for the resume-scoring and Savos assessment models stating which candidate attributes are and are not used as features, pursue ISO 42001 to sit alongside the existing ISO 27001 family, add an EU AI Act deployer pack with an Article 27 FRIA template and instructions-for-use, document concrete human-oversight controls (recruiter override, score-level explanations, per-jurisdiction toggles) rather than the current decision-support framing, and stand up a public trust surface with a status page, model-update changelog and a security.txt or responsible-disclosure contact. Fixing the two regional privacy policies that currently return HTTP 404 is a quick baseline correction.