← all articles in English

Oracle and OpenAI want to make ChatGPT the gateway to enterprise work

The partnership announced at Oracle AI World brings artificial intelligence closer to the systems that run companies. For entrepreneurs, it changes both how work gets done and the discussion about where to build a business.

Two people speak from armchairs on stage in front of a blue screen.
A conversation on the Oracle AI World stage. Image provided by Rodrigo Terron.

In my first article about Oracle AI World, I wrote about a change I consider important: software is starting to do the work, beyond providing tools for someone else to do it.

The October 27 announcements added another question to that discussion.

If artificial intelligence can coordinate activities across different systems, which interface will we use to work?

Oracle and OpenAI announced an expansion of their partnership to bring ChatGPT and Codex closer to Oracle's applications, data and infrastructure. The proposal spans four areas: an integrated workspace, application development, security and a bundle for small businesses. Oracle's official announcement.

What catches my attention is the possibility of changing how people relate to enterprise systems.

For decades, we have learned to work by opening programs, navigating screens and figuring out where each activity should happen.

Now, companies are trying to build an experience where the person starts with the objective.

A conversation could become the starting point for work

One of the announced initiatives is Oracle AI Workspace, being developed to bring Oracle applications and data closer to ChatGPT.

OpenAI gives an example: someone who has just joined the finance team could investigate costs, understand purchasing decisions and identify outstanding month-end closing tasks through an experience that brings together information from different applications.

The proposal preserves the organization's permissions and approvals. Connecting systems therefore does not mean allowing anyone to access any information or execute any action. OpenAI's official announcement.

Think about what this could mean for someone who understands the business but has not mastered all the tools the company uses.

Today, part of the work consists of finding out where information is stored, who can access it and which sequence of screens leads to the desired result.

If AI can coordinate that navigation, a detailed knowledge of every interface may become less important for some activities.

Knowing how to frame the problem, evaluate the response and decide what can be executed becomes even more valuable.

The systems remain necessary. They hold the records, rules and much of the operation. What may change is how we reach them.

Attendees walk through the exhibition area, with an Oracle Fusion Cloud and Industry Applications sign in the background.
Oracle enterprise applications exhibition area. Image provided by Rodrigo Terron.

Creating an app is only part of the problem

Another announcement was Oracle App Composer, a service for creating, publishing and sharing applications with managed operations on Oracle infrastructure.

The proposal allows teams to use tools such as Codex to build applications while IT controls access, data connections, policies and costs. According to Oracle, the service is in limited availability. App Composer details.

This announcement touches on a question that comes up increasingly often when we discuss development with AI.

Someone creates a useful tool for their own work. Colleagues start using it. At some point, that tool becomes part of the operation.

Who looks after it when something goes wrong? Who can access the data? How do you update it without interrupting the team's work?

The easier it becomes to create software, the more people can participate in building it. But the responsibilities of keeping an application running still exist.

To me, there is an important opportunity in that transition: turning an individual solution into something an organization can use with confidence.

This also helps explain why infrastructure platforms and enterprise systems are moving closer to AI coding tools.

Delegating work requires defining authority

The third area is Oracle AI Security Mesh, announced to help find and fix vulnerabilities in applications and infrastructure using OpenAI's security models.

Oracle describes testing within a customer-authorized scope, evidence to assess the issues and approval before fixes are applied. Security announcement.

That detail deserves attention.

When AI only produces text, someone can review it before using it. When it starts changing a system or participating in a financial process, we need to define much more clearly what it has the authority to do.

Who authorized it? What is the limit? Who monitors it? How do you stop or correct an action?

For anyone building products with agents, these questions are part of the product from the start.

A demonstration can impress with its autonomy. A company needs to understand under which conditions it can trust it.

Where does that leave startups?

The partnership also includes Oracle AI SMB Bundle, which combines tools for developing products and running a small business.

The offer announces US$12,000 a year in infrastructure credits, with renewal possibilities and specific conditions. There is an important consideration for readers in Brazil: the published rules require an eligible U.S. business. The components have their own limits and terms. Oracle's published rules.

Beyond the commercial benefit, I am interested in the direction of this movement.

Large platforms want to participate in several stages of a company's journey: helping it plan, build its product, launch it and run the business.

This can make a founder's life much easier. It also makes understanding where your startup creates value even more important.

If an activity becomes available within the tools a customer already uses, why would they buy a separate solution?

The answer may lie in industry knowledge, integration across vendors, the ability to handle exceptions or a deliverable the customer can measure.

There is still room. But we need to look carefully at what makes the solution necessary.

The question now goes beyond software

We are still talking about announcements and products at different stages of availability. An integration presented at an event does not prove that any company can already use it throughout its operation.

Even so, the direction matters.

In the first article, the question was which work a startup could do better.

After these announcements, I would add another:

If customers can request that work directly in the tool they already use, what will make them choose your company?

It is a question about product, distribution and trust.

And it is worth putting on the table while we are building.

Explore an editorial selection
Read the full biographyCollapse biography

Founder of NewHack and an OpenAI Codex Ambassador. Cofounder and former CEO of Shawee, which merged with Rocketseat, where he served as COO and CEO. He applies this experience to connecting companies, developers and entrepreneurs. He also leads AI House Brasil and writes about entrepreneurship, innovation and applied AI for people who build.

WhatsApp