Dubai AI Firm AHOY Labs Ranks Among Global Leaders, Surpassing Major Tech Rivals in Enterprise Benchmark

A Dubai-founded artificial intelligence company has secured a place among the leading performers in an international assessment of enterprise AI systems, recording higher scores than several well-known technology companies in tests designed to measure information retrieval and accuracy.

AHOY Labs placed eighth among 26 systems evaluated through EnterpriseRAG-Bench, an open-source benchmark that examines how effectively AI platforms locate relevant information, produce accurate answers and verify details drawn from large collections of business records.

According to the published results, the company’s platform outperformed competing offerings from OpenAI, Amazon, Microsoft, Google and NVIDIA in the assessment. The results highlight the growing competition in enterprise AI, where businesses increasingly require systems capable of searching internal records without introducing irrelevant information or presenting unsupported answers.

The benchmark evaluates how AI tools perform when confronted with complex corporate information. Unlike systems designed primarily for general conversations, enterprise retrieval platforms must find specific details across extensive collections of documents, messages and other records before generating a response.

AHOY Labs achieved an overall score of 74.95, with 79.6 per cent of questions answered correctly. OpenAI’s File Search recorded an overall score of 61.03 and a correct-answer rate of 69.8 per cent. Microsoft’s Azure AI Search received 48.42, while Google’s Vertex AI Search scored 41.87.

The results also highlighted a difference in the number of irrelevant documents returned by the competing systems. AHOY Labs reported an average of 1.03 irrelevant documents per response, compared with 15.7 for OpenAI File Search.

For businesses evaluating AI solutions, these measures are important because the usefulness of a system depends not only on its ability to produce an answer but also on whether that answer is supported by the right information.

What the EnterpriseRAG-Bench assessment measures

EnterpriseRAG-Bench is an open-source evaluation framework developed by Onyx to assess the performance of enterprise retrieval-augmented generation systems, commonly known as RAG systems.

These platforms combine information retrieval with generative AI. They first search a collection of records for relevant material and then use the retrieved information to formulate an answer to the user’s question.

The approach is increasingly important for organisations that want employees to access information stored across internal databases, documents, communication platforms and business applications.

For example, an employee might need to locate a decision recorded in an old meeting transcript, identify a change discussed in an email thread or compare information contained in several internal reports. A capable enterprise AI platform should retrieve the relevant records and provide a response that accurately reflects their contents.

However, these tasks can become difficult when information is distributed across multiple sources, when documents contain contradictory details or when the available records do not provide a reliable answer.

EnterpriseRAG-Bench is designed to examine these challenges rather than simply testing whether an AI system can generate fluent text.

The assessment uses a simulated corporate archive containing 500,000 records. The collection represents the type of extensive information environment that businesses may need to search when using AI for internal knowledge management.

Questions in the benchmark require systems to identify relevant information across this archive. Some tasks involve details scattered among multiple documents, while others introduce conflicting records or situations in which the correct response cannot be established from the available information.

This makes the evaluation relevant to businesses considering AI tools for complex information retrieval, particularly where the accuracy of an answer can influence operational decisions.

AHOY Labs records a score of 74.95

The published leaderboard places AHOY Labs eighth among the 26 systems included in the assessment.

Its overall score of 74.95 was higher than the results reported for several major technology providers participating in the comparison.

OpenAI File Search recorded 61.03, Microsoft’s Azure AI Search achieved 48.42, and Google’s Vertex AI Search received 41.87.

AHOY Labs also reported that its system answered 79.6 per cent of the benchmark’s questions correctly, compared with 69.8 per cent for OpenAI File Search.

These figures suggest that AHOY Labs performed more strongly than those particular systems on the measured tasks in this evaluation.

However, the results should be interpreted within the scope of the benchmark. An individual assessment measures performance under its own testing conditions, and the rankings do not necessarily establish that one platform is superior for every business application or AI task.

Enterprise AI products can differ in their search architecture, configuration, integration options and intended use cases. Their practical performance may also depend on the quality of the underlying data, the complexity of the organisation’s information systems and the way a platform is deployed.

Even so, a strong showing in an independent, publicly accessible benchmark can provide useful information for organisations comparing enterprise search technologies.

For AHOY Labs, the result offers an opportunity to demonstrate the performance of a platform developed for businesses that need to retrieve information accurately from large internal datasets.

Why precision matters in enterprise AI

One of the most notable aspects of the assessment was the difference in irrelevant-document retrieval.

AHOY Labs returned an average of 1.03 irrelevant documents per response, while OpenAI File Search returned an average of 15.7, according to the published figures.

This measure relates to precision, which describes how much of the information retrieved by a system is relevant to the task.

In enterprise environments, returning large numbers of unrelated documents can create problems even when a system eventually identifies the correct information. Employees may need to spend additional time checking results, separating useful records from irrelevant material and verifying the evidence behind an answer.

A search platform that retrieves fewer irrelevant documents can make it easier for users to identify the information they need.

Precision is particularly important when employees work with extensive archives containing contracts, policy documents, technical records, customer correspondence and internal discussions. These collections may include similar terminology across unrelated subjects, making it difficult for a search system to identify the most relevant material.

A system must also distinguish between documents that merely mention a topic and those that directly answer a question.

For example, a question about a specific contractual obligation may produce numerous documents referring to the same agreement. Only some of those records may contain the relevant clause or the latest applicable terms.

If the search system retrieves too many unrelated records, employees may struggle to determine which information supports the final answer.

The figures reported by AHOY Labs indicate that its platform performed strongly on this aspect of the benchmark compared with OpenAI File Search.

Nevertheless, the number of irrelevant documents is only one measure of performance. Organisations should also examine answer correctness, source attribution, handling of missing information and the consistency of results across different types of queries.

Handling conflicting information and unanswered questions

Corporate archives are rarely perfectly organised collections of consistent information.

A business may have several versions of a policy, email conversations that reflect different stages of a decision, meeting notes that record competing proposals and documents that have not been updated after a change.

An enterprise AI platform must be able to navigate these conditions without treating every retrieved statement as equally reliable.

The EnterpriseRAG-Bench assessment includes scenarios involving conflicting records and questions for which the archive does not contain an answer.

These scenarios are important because an AI system should not simply generate a plausible response whenever a user asks a question. In some situations, the appropriate response is to explain that the available evidence is insufficient.

This capability is particularly relevant in sectors where employees use internal information to support decisions involving compliance, operations, finance or public services.

If a platform produces an answer that is not supported by the underlying documents, users may make decisions based on inaccurate information. A system that recognises gaps in the available evidence can help reduce this risk, although human review and appropriate governance remain necessary.

The benchmark’s emphasis on retrieval quality and answer accuracy therefore addresses a central challenge facing organisations that want to use generative AI with their own information.

For businesses, the objective is not simply to deploy a system that can communicate naturally. It is to ensure that the information it provides is relevant, traceable and consistent with the records available.

A focus on data sovereignty and offline deployment

Beyond the benchmark results, AHOY Labs is positioning its technology around data sovereignty and customer control.

The company says its platform is designed to operate entirely within a customer’s own infrastructure and can function without an internet connection. It also states that its internal tests produced the same benchmark performance in fully offline environments.

This deployment model may appeal to organisations that face strict requirements concerning where information is stored, how it is processed and who can access it.

Data sovereignty refers broadly to the control and governance of data under applicable legal and organisational requirements. For some organisations, meeting these requirements may mean limiting the transfer of sensitive information to external cloud services or ensuring that systems operate within approved environments.

Government bodies, defence organisations, healthcare providers and energy companies can handle information that requires particularly careful management. Their security requirements may affect which AI systems they can adopt and how those systems can be integrated into existing operations.

A platform that can be deployed locally and operate offline may provide an additional option for organisations that cannot rely on continuous connectivity to external services.

However, offline operation alone does not guarantee security or regulatory compliance. Organisations must still evaluate access controls, encryption, system updates, audit capabilities, model governance and the security of their own infrastructure.

They must also determine whether the technology meets the requirements applicable to their industry and jurisdiction.

AHOY Labs argues that its deployment approach enables customers to retain control of sensitive information while using AI to search their internal records.

The company also describes its platform as the highest-ranked sovereign AI system on the EnterpriseRAG-Bench leaderboard. That characterisation reflects the company’s positioning of its product and should be assessed against the benchmark’s definitions and the capabilities of other systems.

Why the results matter for businesses

Enterprise AI adoption is increasingly focused on practical business applications rather than experimentation alone.

Organisations want systems that can help employees locate information, reduce time spent searching through records and make internal knowledge more accessible.

These needs are particularly relevant to large companies with extensive document archives and information spread across multiple departments or applications.

A well-designed retrieval system can help employees find the material they need without manually searching through thousands of records. It can also make institutional knowledge easier to access when staff change roles or when important decisions are recorded across different communication channels.

However, these benefits depend on the reliability of the system.

If a platform repeatedly retrieves irrelevant material or produces answers that cannot be verified, employees may lose confidence in its usefulness. In some cases, they may need to spend almost as much time checking its responses as they would have spent searching manually.

Accuracy and precision are therefore central considerations when businesses evaluate enterprise AI tools.

The AHOY Labs results offer one data point for organisations examining these issues. The company’s performance in the benchmark suggests that its system was able to answer a relatively high proportion of questions correctly while limiting irrelevant retrieval under the tested conditions.

For companies with strict data-handling requirements, the claimed ability to operate on local infrastructure may also be relevant.

Nevertheless, selecting an enterprise AI platform requires a broader assessment than comparing leaderboard scores. Businesses should test systems against their own documents, evaluate how they handle confidential information and determine whether the technology integrates effectively with their existing processes.

Factors such as operating costs, implementation requirements, maintenance, scalability and user experience can also influence whether a platform is suitable for a particular organisation.

Dubai’s growing role in artificial intelligence

The results also place a Dubai-founded company within the wider international discussion about enterprise AI development.

The sector has attracted investment and competition from established technology groups as well as smaller firms building specialised systems for particular business needs.

While major providers offer broad AI platforms and cloud-based services, specialised companies are exploring approaches tailored to particular requirements, including private deployment, information retrieval and data governance.

AHOY Labs’ benchmark performance provides an example of a regional technology company competing in this environment.

For Dubai and the wider UAE, the development of homegrown AI businesses is relevant to efforts to expand the technology sector and encourage innovation across industries.

However, the significance of a single benchmark result should not be overstated. Long-term success in enterprise AI also depends on customer adoption, technical reliability, commercial performance and the ability to meet changing security and regulatory requirements.

A strong evaluation result can help a company demonstrate its capabilities, but businesses will ultimately judge a product on how well it performs in real operational environments.

What the benchmark result means — and what it does not

AHOY Labs’ eighth-place ranking and reported score of 74.95 indicate that the platform performed well within the EnterpriseRAG-Bench assessment.

Its reported correct-answer rate of 79.6 per cent exceeded the 69.8 per cent recorded by OpenAI File Search in the same published comparison. The company also reported a substantially lower average number of irrelevant documents per response.

These results are useful for understanding how the tested systems performed under the benchmark’s specific conditions.

They do not, however, establish that AHOY Labs outperforms every product from OpenAI or all competing products from Amazon, Microsoft, Google and NVIDIA in every category of artificial intelligence.

The assessment concerns enterprise information retrieval and related answer-generation tasks. It should not be interpreted as a universal ranking of general-purpose AI models, coding tools, image generators or other unrelated capabilities.

Similarly, a result on a simulated corporate archive may not perfectly predict how a system performs on every company’s real-world data.

Organisations considering the technology should review the published methodology and test candidate platforms against their own information needs.

The road ahead for enterprise AI

As businesses expand their use of artificial intelligence, the ability to retrieve reliable information from internal archives is likely to remain an important area of development.

Companies need systems that can handle large datasets, identify the most relevant records and distinguish between supported answers and situations in which information is missing or contradictory.

They also need deployment options that reflect their security requirements and operational constraints.

AHOY Labs’ performance in EnterpriseRAG-Bench highlights its position in this specialised area of AI development. Its reported results, combined with its focus on local and offline deployment, provide a basis for further evaluation by organisations looking for alternatives in enterprise search.

The broader lesson is that enterprise AI performance involves more than the ability to generate convincing text. Systems must retrieve the right information, use it accurately and provide results that employees can verify.

As competition intensifies, independent benchmarks can help businesses compare technologies, identify strengths and weaknesses, and make more informed decisions about adoption.

For AHOY Labs, the latest ranking represents a notable result in its effort to establish itself in the global enterprise AI market. For businesses, it offers another platform to consider as they assess the growing range of tools available for searching, analysing and managing corporate knowledge.