Harri

Company overview

Harri (Harri US LLC, 665 Broadway, Suite 402, New York) is a vertically focused HCM and workforce-management platform for service industries — quick-service and full-service restaurants, hotels, grocery and the care sector. Its stack spans talent acquisition (custom career portal, applicant tracking, Talent Scout, talent pools, and the Carri conversational hiring assistant), workforce management and scheduling, CoreHR, employee engagement, insights and analytics, and an 'Agentic AI' layer branded Salli that surfaces talent-gap, compliance and overtime-cost alerts to managers. The company is led by CEO Luke Fryer, a former multi-unit restaurant operator, and reports roughly 60,000 customer locations and 6 million platform users, with offices in New York, London, Hyderabad and Ramallah.

Regulatory exposure

Carri performs applicant screening and interview coordination for high-volume hourly hiring, which puts Harri's pre-hire stack squarely in AEDT territory under NYC Local Law 144 — notable given Harri is New York-headquartered and counts NYC multi-unit brands among its customers — and in Annex III high-risk territory under the EU AI Act for its UK/EU deployments. Harri's own UK master agreement names Regulation (EU) 2024/1689 explicitly, prohibits Article 5 practices, references Article 27 fundamental rights impact assessments, and requires customers to inform affected workers and workers' representatives before subjecting them to the AI Services. But that acknowledgment is entirely contractual and allocative: Harri publishes no bias audit, no model or system documentation, and no NYC LL 144 material, while its MSA places 'sole responsibility' for employment decisions and for AEDT/AI-law compliance on the customer. Illinois AIVI (820 ILCS 42) does not clearly bite, since no video-interview analysis was found, but Illinois HB 3773 and Colorado SB 24-205 duties would land on employers with very little vendor-supplied evidence to lean on.

Path to a higher score

The single largest gain would be commissioning and publicly posting a NYC LL 144 bias audit of Carri's screening logic by a named independent auditor, with selection rates and impact ratios downloadable rather than gated. Beyond that: publish a Carri/Salli system card covering inputs, outputs, intended use, known limitations and measured performance to give Article 11 substance; stand up a real trust or security page that exposes the ISO 27001 and SOC 2 reports (currently only footer badges) alongside a security contact and vulnerability-disclosure policy; and convert the AI Terms Addendum's deployer obligations into customer-facing artifacts — an Article 27 FRIA template, a NYC LL 144 notice pack, and documentation of in-product override, explanation and audit-log controls. Restoring a public help centre with AI-specific admin documentation would also lift customer documentation and human-oversight scores together.