Enterprise search is taking on a new role as AI applications move beyond chatbots and toward systems that retrieve information, reason over it and take action. Tencent Cloud and Elastic are expanding their strategic collaboration around that shift, launching Tencent Cloud Elasticsearch Service (ES) Enterprise Edition to give enterprises a search and retrieval layer for AI applications, enterprise data and agentic workflows.
Search used to be primarily about helping employees find a document, customers locate a product or engineers troubleshoot a system. The rise of generative AI is turning it into something more fundamental: the infrastructure that supplies models and AI agents with enterprise context.
That is the premise behind an expanded strategic collaboration between Tencent Cloud and Elastic, announced in Shenzhen. The companies are broadening their technical and product work across AI search, enterprise data and agentic applications while jointly launching Tencent Cloud Elasticsearch Service (ES) Enterprise Edition.
The new offering brings Elastic’s enterprise search and AI capabilities into Tencent Cloud’s Elasticsearch service, which the company has operated since 2016.
The timing reflects a larger change in enterprise AI architecture. Large language models (LLMs) are trained on broad datasets, but they generally do not have direct knowledge of an individual company’s latest documents, transactions, product information, security events or internal processes. To answer business-specific questions reliably, AI applications need a mechanism for retrieving relevant information at inference time.
That is where search enters the AI stack.
In a retrieval-augmented generation (RAG) architecture, a search system retrieves relevant enterprise content before an LLM generates an answer. Agentic applications extend that concept further: an AI agent may need to repeatedly locate documents, records, tools or other information while completing a multi-step task.
Elasticsearch is particularly suited to this transition because it can combine traditional keyword retrieval with semantic search. Keyword search remains useful when users need exact matches, identifiers or technical terms, while vector and semantic retrieval can identify information based on meaning.
The enterprise challenge is increasingly to combine both.
Tencent Cloud ES Enterprise Edition allows organizations to use keyword and semantic search together, while enabling LLMs and AI agents to directly consume search results. The company also says enterprises can maintain searchable historical data, giving AI applications access to a wider body of business context.
That makes the product less about “search” in the traditional user-interface sense and more about context infrastructure for AI.
The distinction is important as enterprises evaluate competing approaches from cloud providers and AI platform vendors. Microsoft, Google Cloud and Amazon Web Services all provide technologies for building RAG applications, vector search and AI agents. Databases and data platforms are also adding native vector and semantic capabilities.
Elastic’s positioning is different: it comes from a search technology stack already used for applications ranging from e-commerce product discovery to application search, observability and security analytics. Elastic says its Search AI Platform is used by thousands of companies, including more than half of the Fortune 500.
Tencent Cloud adds another dimension: operating search infrastructure at large scale within China’s cloud market and connecting it to the company’s broader AI ecosystem, including Tencent Hunyuan large language models and the TCRay inference platform.
The companies say Tencent Cloud’s engineering work has produced improvements in selected AI-search workloads, including up to fivefold performance gains, a 60% reduction in hybrid-search latency and more than 50% lower memory consumption.
Those figures are vendor-reported rather than independent benchmarks, but the underlying optimization problem is significant. AI applications can generate enormous retrieval workloads, particularly when agents repeatedly search enterprise repositories during long-running tasks. Retrieval latency and infrastructure consumption can consequently affect the cost and responsiveness of an AI application.
Tencent Cloud says its Elasticsearch service now operates 20,000 clusters and 100,000 nodes, with the infrastructure supporting traffic peaks associated with large events including the China Media Group Spring Festival Gala and the Paris Olympic Games.
Customer deployments illustrate the intended use cases.
At Tencent ima, ES serves as the retrieval layer for a knowledge base supporting keyword and semantic search. Tencent Cloud says optimization for concurrent knowledge-base queries reduced memory consumption by 71% and improved retrieval performance by 58%.
At electric vehicle manufacturer NIO, Elasticsearch forms part of a security data intelligence infrastructure. The system processes hundreds of billions of security records in a quarter and uses search to identify information relevant to an event before AI performs additional analysis. According to Tencent Cloud, an attack chain that previously required most of a day for manual investigation could be retrieved and reconstructed in approximately two minutes.
These examples point toward one of the more consequential developments in enterprise AI: search is becoming an intermediary between data and intelligence.
Elastic’s Han Xiao describes the changing role of search as “test-time compute”—the idea that improving an AI system does not always require a larger model or additional training data. Better retrieval and additional inference can give an existing model more relevant information when it needs to make a decision.
That concept could become increasingly important as AI agents take on longer-running tasks.
An agent researching a customer issue, investigating a security incident or assembling a business report may perform dozens of retrieval operations rather than a single search. In those environments, the quality, speed and freshness of the retrieval layer directly influence the agent’s downstream reasoning.
For enterprise IT teams, that creates a new architectural consideration. Selecting an LLM is only one part of an AI deployment. Organizations also need to determine how enterprise information will be indexed, retrieved, secured, updated and delivered to models and agents.
Tencent Cloud and Elastic are betting that search will occupy that critical layer.
Market Landscape
The enterprise search market is converging with RAG, vector databases, AI infrastructure and agent platforms.
Historically, Elasticsearch competed primarily with search and observability technologies. Today, its competitive environment increasingly includes cloud-native vector databases, database vendors adding vector search and AI platforms from Microsoft, Google and Amazon.
The emerging differentiation is less about whether a platform supports vector search and more about how effectively it handles hybrid retrieval, enterprise-scale indexing, security, latency, integration with models and agent connectivity.
This matters because retrieval quality can become a bottleneck even when the underlying LLM is highly capable. An AI agent cannot reason effectively over information it cannot find, and an enterprise cannot easily trust answers grounded in stale or incomplete data.
Tencent Cloud’s scale gives the partnership a particularly significant regional dimension. Its integration with Hunyuan and TCRay also creates a vertically connected stack spanning cloud infrastructure, search, inference and AI models.
For enterprises already operating Elasticsearch environments, the ability to add AI search without fundamentally redesigning existing data architecture could be one of the offering’s more practical advantages. Tencent Cloud is also offering migration and upgrade services for customers using open-source editions or self-managed clusters.
The broader direction is clear: enterprise search is evolving from an employee-facing discovery tool into a machine-facing context layer for AI systems.
Top Insights
- Tencent Cloud and Elastic launched ES Enterprise Edition, combining hybrid search and AI capabilities to provide enterprise data context for LLMs and agents.
- AI agents increase retrieval demands because autonomous systems must repeatedly locate current enterprise information, tools and records during multi-step workflows.
- Tencent Cloud’s engineering optimizations reportedly deliver major gains in selected AI-search workloads, addressing latency, memory consumption and large-scale retrieval costs.
- NIO’s security deployment demonstrates how enterprise search can help AI reconstruct complex attack chains from hundreds of billions of security records.
- The partnership strengthens Tencent’s AI ecosystem, connecting Elasticsearch with Hunyuan models and TCRay inference while expanding Elastic’s reach in China’s cloud market.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
