My Top 5 Key Takeaways from Oracle AI World 2025
Last week I attended Oracle AI World in Las Vegas (Oct 13–16). Until just a couple of months ago, the event was still called CloudWorld, but the new name proved fitting—the conference truly lived up to its AI focus.
It perhaps can be said that Oracle was late to the Cloud game (although paradoxically this lateness helped them build a new modern Cloud infrastructure from the ground up). After attending AI World, it is very clear to me that Oracle does not intend to be late to the AI game – and they are not – Oracle Cloud Infrastructure (OCI) is already used to train some of the largest Language Models (LLMs) in the industry.
Below are my key takeaways from the 2025 Oracle AI World conference:
1. AI Changes Everything
This was Oracle’s theme and slogan for the conference. It is a sentiment that most of us are already familiar with. However, most of us, if we are honest, don’t know exactly how that is going to happen in real, day-to-day practical terms. Well, after attending AI World I have a much better idea of how AI will ‘seep into’ every corner of day-to-day work life and be something that we use all the time – whether we are aware of it or not. Oracle is in a somewhat unique position within the large Cloud providers in that they have a full suite of Oracle enterprise business applications running in their Cloud in addition to having all the platforms, services and tools to enable creation of custom Cloud applications. So they were able to paint a picture of how the average person going into work over the next several years will be engaging with AI. The opportunities for efficiency, productivity, and innovation were self-evident and exciting. Every layer of the Oracle platform and product suite has been infused with AI capabilities including:
- SaaS (Fusion Applications)
- Infrastructure
- Integration, ETL and Data Pipelines
- Data
- Analytics and Reporting
2. Oracle Announces Oracle AI Database 26ai– Much More Than Just a Database
This isn’t the Oracle database you grew up with. Frankly it is almost hard to just call it a database. With all the features and capabilities that are now part of the baseline Oracle AI Database 26ai it is more like a data ecosystem all on its own. The world of data and what we want to do with data has been changing for a while but when ChatGPT came out in Nov 2022 and the world was introduced to LLM’s and what they could do, everything changed. Now everyone was aware of the immense power of all types of data if they could be harnessed and used. Oracle has responded to this by significantly expanding the types of data that can be stored/managed within the Oracle database, providing capabilities in the database itself to leverage AI and integrating the database with modern AI models and frameworks. Below is a summary of the capabilities of the new Oracle AI Database 26ai (not intended to be an exhaustive list):
- Supports all data types (Relational, JSON, Spatial, Graph, AI Vector, Media) and all types of data (unstructured, structured and semi structured)
- Ability to call external LLMs and embedding models
- Ability to vectorize your data, store it in Oracle Database 26ai and conduct similarity search – in conjunction with simultaneous search of traditional relational data (if desired)
- Built in Retrieval Augmented Generation (RAG) capabilities
- Built in machine learning model creation, evaluation and deployment with Oracle Machine Learning (OML)
- New ability to extensively annotate the database and its schemas to aid in LLM understanding
- New built-in capability to build, deploy and manage in-database AI agents
- Integration with popular Agentic AI frameworks (e.g., GCP Vertex AI, Bedrock, LangGraph, CrewAI, etc)
- Support for Apache Iceberg – Oracle Database 26ai can read and write to Apache Iceberg – underlying data does not have to be in OCI Object Storage – can be in non-Oracle object storage (AWS S3, ADLS, GCS, etc)
- Support for a full medallion architecture (Bronze, Silver and Gold layers – note: Oracle recommends that the Gold Layer should be a full instance of Oracle Autonomous Data Warehouse – although it does not have to be)
- Includes an ‘Autonomous Database Catalog’ that can act as a ‘federated catalog of catalogs’ identifying data and objects in non-Oracle data stores (e.g., AWS S3, Azure, Oracle, GCP, Databricks, etc).
- Oracle AI Database 26ai can run natively in AWS, Google and Azure (known as Oracle Database@AWS, Azure, GCP)
- SQL Developer Assistant enables the creation of database schemas using natural language
- Includes a capability to generate synthetic test data
- Via APEX (which is part of Oracle AI Database 26ai), you can now describe what you want an application to do in natural language and Apex will generate a human readable Open Application Specification of the application
A couple important points to note:
- It should be noted that Oracle AI Database 26ai still comes in two flavors: 1. Autonomous Transaction Processing (ATP) for transaction processing (data is stored in rows) and 2. Autonomous Data Warehouse (ADW) (although that name has been changed-see below) for AI and analytics use cases (data is stored in columns).
- The new name for Autonomous Data Warehouse (ADW) is Autonomous AI Lakehouse. This may not be the official Oracle explanation but this is effectively what has occurred. So when you see Autonomous AI Lakehouse you should think ADW but with the very important distinction of it now being multi-cloud and open through support for Apache Iceberg and Delta, in addition to Autonomous AI Lakehouse having all the new capabilities of Oracle AI Database 26ai.
3. Oracle Announces AI Data Platform
This will be a critical platform for Oracle going forward and they are investing a lot of money into it. It is a unifying layer that enables a company to manage and get visibility into all its data as well as rapidly build and deploy AI Agents and AI applications of all types.
Here are some of the important features and things to know about Oracle’s new AI Data Platform (AIDP):
- AIDP unifies access to all your data—inside or outside the Oracle Cloud—so you can catalog, govern, and build AI apps from a single place
- Enterprise catalog is central to AIDP’s value proposition- the catalog will help organizations manage and get visibility into their entire data estate – no matter where that data resides – in the Oracle Cloud or not. And it is not just a catalog for data, datasets and reports – it will also be a catalog of AI agents – no matter where those agents reside
- AIDP is tightly integrated with Oracle AI Database 26ai
- Can be used to rapidly design, build and deploy lakehouses leveraging a medallion architecture (i.e,. bronze, silver and gold layers)
- Supports Apache Iceberg and Delta files/tables
- Supports multiple LLMs and industry leading AI models and frameworks – including open source
- Provides an integrated, managed notebook-based developer experience to build AI applications; supports Native SQL, Python, Scala and Java; includes Git integration
- Supports Apache Spark and Flink engines
- Scales elastically across CPUs, GPUs and distributed clusters
- Provides a single development environment for data integration, data engineering and data science
- Makes it easy to build and deploy RAG applications and GenAI chatbots
- Enables zero copy access to data in Oracle Fusion Data Intelligence (FDI) – so you could use AIDP’s AI capabilities to build an AI application using data from FDI without moving data out of FDI
- By the end of the year it will include an Agent Hub which will enable all types of AI Agents to be developed in a no code/low code manner
- Includes comprehensive and integrated security, governance and lineage
- Oracle is positioning AIDP to be very price competitive with other Clouds (i.e., all things being equal Oracle is positioning AIDP to be a significantly less expensive alternative)
Learn about how Apps Associates has used AI Data Platform:
4. Agents Rule
If you have not heard yet the world is about to be overrun by an army of agents that will do everything for us in a timely and error free manner. Ok I am being a little tongue-in-cheek here but that does not change the fact that AI Agents can be tremendously powerful and hold out immense promise for improving productivity and efficiency and making it a lot easier to get answers to business questions.
At AI World it was great to hear Oracle speaking about AI Agents ‘augmenting’ humans rather than replacing humans. In case you are not familiar with AI Agents, in layman’s terms, AI Agents can ‘do things on their own’. AI agents can reason through interactions with one or more large language models (LLMs), make decisions, and autonomously take the next step in a process. They can participate in workflows, access data and tools (including via APIs or the internet), and return results or actions to humans or other systems — all while operating within defined policies, permissions, and credentials. AI Agents are already and will continue to be tremendously important as organizations embrace AI. Throughout AI World, Oracle emphasized the ability to create AI Agents in a user-friendly, low code/no code manner. Oracle has provided a number of places where AI agents can be developed in the Oracle Cloud. When to use which platform depends on the use case and is outside the scope of this article. Oracle enables AI Agent development via these platforms
- AI Agent Studio in Fusion SaaS apps
- Agentic AI capabilities in Oracle AI Database 26ai
- Agent Hub in Oracle AI Data Platform
- OCI Gen AI Service
5. Natural Language Query (NLQ) is Evolving and Improving
For those that are not aware, NLQ is when you ask a question in your natural language (e.g., English, Spanish, Chinese, etc.) and you receive the answer back without writing any code or creating a report of any kind. As you would suspect, an LLM is used to interpret the natural language question, generate the SQL (or whatever language) against the database and return the results to the user – which could include automatically generating a report or visualization. If you have ever used ChatGPT or Claude or another AI, you know that they can sometimes ‘hallucinate’. This can be very dangerous if you are asking about last quarter’s financial results – you need an exact answer that you can trust. Along these lines, one of the product leaders at Oracle has a famous saying that has been making the rounds – “GenAI is magical but it is not magic”.
That being said, the ability to simply ask questions and get answers related to your data, financial results, business operations, products, customers, warehouses, etc is the most natural interface and is the one that is likely to be the most predominant in years to come.
At AI World, I saw many demos where a person would type in a question and a report or visualization would be automatically generated with the correct answer. So the technology is there and Oracle is building it into all of their reporting and analytics platforms (e.g., Oracle Fusion Data Intelligence and Oracle Analytics Cloud). Oracle is also taking steps to give users more visibility into how the answer is being derived. But, it should be remembered that it takes some setup, configuration and training of the LLM to be able to get answers that you can trust 100% of the time – regardless of the technology being used.
Another way that Oracle is enhancing the accuracy and security of NLQs, is that Oracle is now supporting Model Context Protocol (MCP). Without getting too technical, MCP keeps the LLM out of the database. It validates the queries the LLM generates and executes them safely to return accurate results.
Closing Thoughts
AI presents us all with an exciting future. Oracle is clearly embracing AI in a creative and aggressive manner across its entire infrastructure, platform capabilities and products. Oracle has also made a significant move to becoming open and supporting a multi-cloud future (e.g., running a Lakehouse on the Oracle Cloud where the underlying data is in AWS S3 or running Oracle AI Database 26ai on another cloud to name a couple examples).
Apps Associates is at the forefront of AI and determining how to leverage AI to achieve strategic goals, improve efficiency, accelerate innovation, enhance the customer experience and leapfrog the competition. We can help you navigate this new world of Oracle AI capabilities either in the Oracle Cloud or other clouds.
Contact Apps Associates today to have a discussion about how to leverage Oracle AI in your organization.
