Jamil Shinawi – Ahoy CEO
The sovereign-by-design AHOY Labs AI platform has surpassed enterprise search systems from OpenAI, Amazon, Microsoft, Google, and NVIDIA on EnterpriseRAG-Bench. The UAE-founded deep-tech firm ranked eighth worldwide among 26 competing systems on the open-source evaluation framework.
EnterpriseRAG-Bench evaluates systems using an archive of 500,000 corporate records, including emails, chat threads, support tickets, sales files, and meeting transcripts. Participating software must answer 500 complex questions where data is distributed across multiple files, contains contradictory statements, or lacks answers entirely.
Benchmark Results for AHOY Labs AI
The benchmark scores systems on response accuracy and precision. AHOY Labs achieved an overall score of 74.95, answering 79.6% of questions correctly while returning 1.03 irrelevant documents per answer. In comparison, OpenAI File Search recorded an overall score of 61.03 with 69.8% accuracy, returning an average of 15.7 irrelevant documents per answer.
Amazon Q scored 48.96 with 55.4% accuracy and 1.49 irrelevant files, while Microsoft Azure AI Search achieved 48.42 with 56.4% accuracy and 3.25 irrelevant files. Google Vertex AI Search scored 41.87 with 49.2% accuracy, and NVIDIA AI Blueprints recorded 37.73 with 59.6% accuracy and 7.72 irrelevant files.
“We built AHOY to prove that sovereignty and world-class performance are not a trade-off. Outperforming the biggest names in technology, on both accuracy and precision, shows that regional organisations can keep full control of their data without settling for second-best AI.”
Jamil Shinawi, Founder and Group CEO of AHOY
Data Sovereignty and On-Premises Architecture
The system operates directly within client infrastructure, functioning on dedicated hardware without requiring external cloud connections. Internal testing confirmed that the deployment delivers identical performance metrics while operating fully offline, supporting cybersecurity and compliance standards in regulated sectors.
Organizations in government, defense, healthcare, and energy require verifiable data privacy controls. Consequently, operating air-gapped systems ensures sensitive institutional records remain strictly within institutional boundaries during real-time retrieval operations.
Enterprise Retrieval Precision and Noise Reduction
Returning fewer irrelevant documents directly reduces operational noise during complex document search tasks. Specifically, the AHOY Labs AI system produced approximately a fifteenth of the irrelevant files generated by OpenAI File Search during the evaluation.
Irrelevant documents increase operational costs, obscure factual evidence, and complicate audit trails. For regulatory bodies and public institutions, retrieving precise source documentation remains essential for verifying institutional automated decisions.
Open-Source Evaluation Framework
EnterpriseRAG-Bench was developed by Onyx under the MIT open-source licence and includes an accompanying technical research paper. Onyx verifies all benchmark submissions before publishing scores on the public leaderboard and excludes its own tools from the rankings.
The underlying infrastructure is powered by AHOY Machine Studio, which provides model governance, development, and operational deployment. The architecture operates across cloud, edge, and on-chip environments in international markets including Saudi Arabia, the UAE, Canada, and the United Kingdom.
