RESEARCH NOTE: Oracle’s Intriguing Approach to AI for App Developers

By Jason Andersen - May 9, 2024

Over the past few days, Oracle has made two AI-related announcements that on their surface may not seem particularly exciting or unique. However, when you peer a bit deeper and consider the motivations behind Oracle’s decisions, these announcements are in fact quite interesting.

Last week, with some degree of fanfare, Oracle announced the general availability of Oracle Database 23 AI. That was followed this week by the company’s announcement that it intends to release an AI-powered development assistant (or “code companion”) called Oracle Code Assist. Before I get into the details, let’s talk about what I believe is driving Oracle’s near-term AI aspirations.

“At Oracle, we put the customer first”

This statement was very clearly made early on in both briefings. Notably, the database briefing was hosted by Larry Ellison himself, who made the statement. At first, it sounds totally hokey. I mean, don’t all companies say they put the customer first? It’s kind of a throwaway sentiment—unless you can back it up.

As I reflected on both briefings, I realized that while Oracle has not gotten the same AI attention as, for instance, Microsoft or Google, the span of its product suite already contains many of the required pieces for AI success: databases, cloud, SaaS, high-end servers, and so on. Plus Oracle is still very much in use across many industries and takes up plenty of attention among, say, the people running datacenters. So while it might not be top-of-mind in the AI media sphere, Oracle remains top-of-mind for enterprise IT people. And both of these releases reflect that point really well.

Incorporating Oracle Data into AI

While the more general-purpose AI models get a ton of attention, enterprises that have bought into AI want responses that are customized to their own data, standards, and policies. They also want all of that content to be secure and timely. Today, a lot of the relevant enterprise data is stored in relational databases, especially Oracle databases. Oracle is responding with support for vector search, which enables users to find what they’re looking for by semantics or meaning versus simply matching. This is the same way that large language models (LLMs) index data for retrieval by AI agents and apps.

In addition to the data now being indexed in a way that is consumable by an LLM, Oracle natively supports retrieval-augmented generation (RAG), a common AI technique that allows the data within a model to be semantically matched with public data. This allows developers to create more meaningful and specific user experiences by leveraging Oracle databases in a direct and secure way.

An Oracle-Optimized AI Assistant

There has been a lot of recent dialogue about AI assistants, and we fully expect that trend to continue. (See for example, the analysis we published last week on the new Amazon Q release.) Alongside customer service chatbots, developer assistance tools are one of the top use cases for generative AI. That’s why it was not surprising that Oracle pre-announced Oracle Code Assist this week (albeit with no specifics yet about when it will actually be released).

While Oracle Code Assist has a similar base set of features as other developer AI tools, it stands out in an interesting way. As the steward of Java programming and the countless databases built on Oracle technology, Oracle has specifically optimized and trained its AI model on those technologies. So, instead of casting a broad net to win over as many hearts and minds as it can, Oracle wants to deliver the best possible experience to its own core set of users. This should lead to many benefits, including massive simplification of upgrades and more precise and consistent suggestions from the AI assistant throughout the entire application development life cycle.

A screenshot of Oracle Code Assist, which is currently in use by internal Oracle engineering teams.

Oracle Seems to Have Met Its Primary Goal

Assuming it delivers on all of this, Oracle’s claim about putting its customers first does ring true. Oracle appears to have made a conscious decision that its enterprise customers need not just an AI strategy, but an AI strategy that leverages their existing investments in Oracle. While this may lack the broad-based appeal of other companies’ AI announcements, these new innovations are intriguing because they are built around creating the best experience for a subset of an enterprise stack.

It would be great to see Oracle provide meaningful benchmarks for performance or efficiency gains that come from using Oracle Code Assist. Benchmarks are something that Oracle has done well with databases and servers over the years, and they would seem to be the best way to validate this approach to AI. There are two ways Oracle could do this. Given that the Oracle internal engineering team is already using Oracle Code Assist today, the company’s internal performance metrics may be a good way to differentiate the product if and when it becomes generally available. That’s one approach. Oracle could also embrace some industry leadership and work with third parties—customers, partners, or even competitors—to develop a sustainable cross-industry benchmark. Either way, as a later entrant to this AI product category with a more focused value proposition, Oracle will have its work cut out for it.

Jason Andersen
+ posts

Jason Andersen is vice president and principal analyst covering application development platforms, technologies, and services. Jason brings over 25 years of experience in product management, product marketing, corporate strategy, sales, and business development at Red Hat, IBM, and Stratus to his work for MI&S and its advisory clients. Working both in the field and in the headquarters of some of the most innovative technology companies, Jason has a wealth of experience in building great products and driving their adoption across a broad spectrum of industries and use cases.