Company Cloud

Change.orgThe customer this was delivered for.Action CenterSave this case, request the same outcome from Deel, share it, or get in touch.

300+ Admin Hours Saved Monthly

Change.org (https://Change.org) faced significant operational complexity due to its distributed workforce across more than 25 countries. They previously managed payroll with countless vendors and multiple platforms, which was time-consuming and prone to errors. Their legacy provider also suffered from a lack of automation, slow support, and a clunky user experience. These issues necessitated a search for a more effective and automated solution. Change.org (https://Change.org) migrated to Deel Payroll, seeking better support and an improved user experience. They appreciated Deel’s immediate support specialist responses, which was a key factor in their decision. The team also valued Deel's advanced and user-friendly platform, contrasting it with outdated systems lacking automation. Deel also provided comprehensive local legal support and compliance handling, automating and customizing solutions for Change.org (https://Change.org)'s global needs. Since implementing Deel Payroll, Change.org (https://Change.org) successfully saved over 300 hours on administrative tasks each month. They also utilized Deel EOR to retain valuable employees requiring visa sponsorship. This allowed team members to live where they preferred, contributing to improved retention. These outcomes empowered Change.org (https://Change.org) with confidence in their future global workforce management.

Delivered by

Deel

Change.orgThe customer this was delivered for.Action CenterSave this case, request the same outcome from Fractional AI, share it, or get in touch.

77% More Content Violations Detected, False Positive Rate Cut in Half

Change.org (https://Change.org) faced the challenge of moderating thousands of daily campaigns to meet community guidelines, requiring a system to manage content violations like hate speech and misleading information. Their existing no-code solution, built with Google Sheets, Zapier, and GPT, automatically flagged about half of content violations but struggled with a 50% false positive rate for petitions sent for manual review. This meant significant human time was spent reviewing harmless content, and many dangerous petitions still fell through the cracks. They aimed to move beyond spreadsheets to a more robust, integrated system that caught more violations and reduced false positives. Change.org (https://Change.org) partnered with Fractional AI to develop a more robust and scalable AI-powered content moderation system. Through over 100 experiments, Fractional AI created an AI system designed to navigate the nuances of content moderation. This new system was integrated directly into Change.org (https://Change.org)'s tech stack, replacing the previous spreadsheet-based workflow with a streamlined REST API workflow. The solution involved fine-tuning GPT 3.5, using the OpenAI API moderation endpoint, and incorporating structured output with chain-of-thought prompting. The implemented AI system significantly improved content moderation, detecting 77% of content violations. It also dramatically reduced the false positive rate, cutting it in half (from 1 in 2 to 1 in 4 petitions flagged for review). This reduction in false positives meant human reviewers spent their time more efficiently. The system replaced a complex spreadsheet workflow with a robust REST API, seamlessly integrating into Change.org (https://Change.org)'s environment, all while keeping daily operational costs around $30.

Delivered by

Fractional AI