Presented by RavenDB
Document databases have long been the backbone of critical applications — now, they’re entering a new era of intelligence and capability. By embedding generative AI directly into the data layer, the impact is not just a technical upgrade — it’s a strategic, transformative shift for faster delivery, leaner operations, and smarter scaling.
“When you can put the entire weight of a large language model inside the database, you can do a lot of things that are flat out amazing,” says Oren Eini, CEO and founder at RavenDB. “It’s a rethinking of where AI belongs in the enterprise software stack. With AI inside the data engine, organizations of every size can easily create intelligent applications. Democratizing the ability to apply AI in all aspects of your system will result in a massive change to how we think about building software.”
Turning queries into intelligent actions
In the past, you needed to be an expert in data tagging, spend enormous time and effort properly cataloguing all your data, create a full search system, and so on. With AI at the data level, you can use that model to do a lot of the work here, and then you can apply vector search on top of that.
Generative AI lets users generate, enrich, classify, and automate content and decisions directly within the database using large language models. And AI capabilities where the data actually lives enables secure, efficient, real-time execution, whether classifying documents, summarizing customer interactions, or automating workflows. It lets teams build features directly from the data they already manage, with no dedicated AI team required.
“That gives you a lot of power, which previously only the largest companies had available,” Eini says. “You can have a really high-quality system, comparable to what Google or Microsoft can run, but in a fraction of their budget and with a swift deployment process. Given your existing data, you can go from nothing to having a first-grade intelligent system within days.”
This kind of integration is huge for organizations that don’t have large AI teams or dedicated MLOps infrastructure. For many purposes, adding generative AI (GenAI) capabilities directly into a core database engine is a leap forward from service wrappers or proprietary cloud stacks, significantly reducing the complexity involved.
It’s a straightforward approach, Eini says, pointing at how RavenDB enables customers to integrate AI in their product. This approach lets users do more with their data, with built-in summarization, classification, and tagging support. RavenDB users can even create workflows that enrich the dataset directly from within the database, and use data to generate additional documents and information – moving beyond passive data retrieval toward using data as a catalyst for innovation.
The RavenDB difference
“In essence, we aim to commoditize the usage of LLMs in your systems by making it easy, predictable, sustainable, and affordable,” Eini says. “Most other AI integrations in other systems simply allow you to make remote calls to the model which is nice if you want to feed some data to the model directly from the database — and then return the response to the user. The problem with that workflow is that it doesn’t allow you to really utilize the outcome of the model’s action, or personalize and optimize the LLM behaviors,” he notes.
“On the other hand,” he explains, “RavenDB’s approach is to create a dedicated pipeline, where the developers can simply input a prompt that tells the model what needs to be done, the model processes the request along with the data in the database, and applies the results to the database on a continuous basis.”
RavenDB will invoke the model to apply the user’s logic to any new data as it’s added, which means no longer waiting a significant amount of time for results. Injecting “intelligence” to the database also prevents utilizing the model if there is no need to do so. RavenDB can detect if that specific data is pertinent to the model’s defined task, and skip the model invocation if it isn’t (eliminating unnecessary model calls and the associated costs).
RavenDB’s new feature supports safely using any LLM directly inside the database, whether it’s a commercial offering — from the likes of Open AI, Mystral, or Grok — or an open source model. Customers can also plug in specialty models, such as medical LLMs, or ones that were fine-tuned for the customer’s specific needs. The “governance as default” approach that RavenDB uses inside its products limits the capabilities of what the model can do, so that it can’t make any unapproved changes to the database, or access internal systems — making application development not only faster but more secure, and and giving risk teams greater peace of mind.
Instead of needing complex data pipelines, or vendor-specific APIs and vast amounts of engineering and data science know-how in the quest to move from prototype to production, teams across the organization can now run GenAI tasks directly inside the database, significantly narrowing the gap between experiment and final result.
“Developers suddenly have complete control over all the things that plague them most, including cost, performance, and compliance,” Eini says. “The key from our perspective is that by making it easy for you to utilize AI in your systems, we make it possible for you to apply sophisticated models everywhere. The idea is that instead of going into months-long integration processes, they’re now as easy as running a query. The end result is a technology that makes going from idea to implementation faster than ever, and nearly seamless.”
Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact
