Oracle Just Declared War on SaaS with a No-Code-to-Pro-Code Agent Builder

TL;DR: Oracle dropped an AI agent builder inside its Fusion suite that lets business users describe agents in plain English and gives devs full Claude Code and Codex access, all targeting Oracle's governed runtime. They're not building agents. They're building an agent runtime with built-in governance, and they're telling SaaS vendors directly: "We're the ones disrupting SaaS." This is the most serious enterprise agent play yet, and it changes what "no-code agent building" actually means.

I arrived at this conclusion slowly, but I'll say it fast: the AI agent contest isn't about who builds the smartest agent. It's about who owns the governed runtime. And the loudest shot fired in that direction so far came last week from the company everyone wrote off as too slow for this era.

Oracle.

On 14 July 2026, Oracle launched an AI-native builder experience inside Oracle AI Agent Studio for Fusion Applications. Everyone's talking about the builder, because it's a properly scary proposition for every SaaS vendor whose strategy was "stick an AI chatbot on top and call it a platform."

The pitch is simple but the architecture is what matters. Business users describe what they want an agent to do in natural language. The system generates a Fusion Agentic Application. Not a standalone bot, not a Zapier workflow with a chatbot slapped on. A governed, auditable application that lives inside Oracle's Fusion Cloud, inherits role-based access controls, executes against real business objects, and logs its actions into the audit trail. Developers who want more control can drop into VS Code with Claude Code, OpenAI Codex, Git-based workflows, and CI/CD pipelines, all targeting the same runtime.

Two tiers, one governed execution environment. That's the bit everyone else is missing.

Why would Oracle of all companies get this right?

I've been tracking enterprise agent builders for 18 months. Microsoft put Copilot Studio inside Microsoft 365. Salesforce has Agentforce. SAP has Joule. They're all building agents, and each now offers no-code builders with pro-code escape hatches. What none of them has done is make the governed, in-system-of-record runtime the centrepiece the way Oracle just did.

Here's the difference. When you build an AI agent in a standalone tool and connect it to your ERP, you have a security problem you didn't have yesterday. That agent needs access to customer records, financial data, HR information. Who gave it that access? Under what policy? Can you see what it did last Tuesday? If it approved a £50,000 purchase order, where's the audit log?

Most agent builders answer those questions with a shrug and a link to their SOC 2 report.

Oracle's answer is different because their agents aren't connecting to Oracle. They're born inside it. They inherit the permissions of the user who configured them. They operate within the same governance framework that controls what your finance team can and can't do. For the no-code agents, data stays inside the tenant and actions are logged through Fusion's existing audit infrastructure. The pro-code path, leaning on Claude Code and Codex, brings external-model questions.

This is not a product feature. This is the product.

Chris Leone, Oracle's EVP of applications development, put it bluntly: "Enterprise software is moving beyond systems that record work to systems that actively drive and execute outcomes." The subtext there is about two pages long. What he's saying is that the SaaS model, where you pay a vendor to record and manage your data while you still do most of the actual work, is the thing being disrupted. And Oracle wants to be the disruptor, not the disrupted.

During analyst briefings they were apparently even more direct. I wasn't in the room, but the line circulating from those sessions is: "We're the ones disrupting SaaS."

That's not marketing bluster. That's a statement of architectural intent.

What does this actually look like to use?

The builder experience has two faces. For the business user, it's a natural language interface: "Create an agent that monitors overdue invoices, prioritises them by amount and customer relationship, drafts collection emails, and routes anything above £100K to the regional controller for manual approval." The system generates the agent team, the workflow, the approval routing, and the execution plan.

For developers, the AI Studio Skill brings the whole thing into tools they already use. VS Code. Git. Claude Code and Codex as coding assistants. Local validation, debugging, CI/CD. They're not learning "the Oracle way." They're building in their normal environment, targeting Oracle's governed runtime.

Oracle introduced 50-plus pre-packaged agents back in 2025, and the AI Agent Marketplace now lists more than 100 partner-built agents across ERP, HCM, SCM, and CX. Over 80,000 certified experts are now trained on the platform. And it's included at no additional cost for existing Fusion customers.

(A quick aside: "no additional cost" for Oracle means something different than it does for, say, Notion. Oracle contracts run well into six figures, often to seven. But if you're already on Fusion, the agent builder doesn't cost extra.)

So where does that leave the rest of us?

Here's where I get personal. I run a no-code platform. We think about governed execution, role-based permissions, and audit trails every single day. That architectural philosophy (agents inherit permissions, data stays in your tenant, every action is traceable) is exactly how we build at Stacker. The difference is we do it for mid-market teams that don't have an Oracle contract, an Oracle implementation partner, and six months of change management ahead of them.

The principle is the same. The price tag and the complexity are not.

Oracle just validated the entire category. They've said, at enterprise scale, with enterprise governance requirements, that the right way to build AI agents is inside the system of record, with inherited permissions and native audit trails. That's the architecture we've been betting on. It just got a very loud endorsement from a company with more than $50 billion in annual revenue.

The practical question for operations leaders is straightforward. Do you need Oracle-scale governance with Oracle-scale complexity? If the answer is yes, Oracle AI Agent Studio is the most complete answer on the market right now. If the answer is "I need governed agents without an eight-figure IT budget," the honest answer is the market is still figuring out how to deliver that architecture at mid-market prices.

The takeaway:

The AI agent market just split into two lanes. Lane one: build smart agents fast, governance optional, hope for the best. Lane two: build governed agents inside the system of record, where permissions, audit trails, and data residency are the foundation, not an afterthought.

Oracle planted a flag in lane two. Loudly. They told every SaaS vendor whose strategy is "we bolted a chatbot onto our API" that their clock is ticking. And they made the governance-first architecture the benchmark that every other agent builder will now be measured against.

For mid-market teams, the question isn't "should we adopt Oracle's approach?" It's "how do we get governed agents without the Oracle overhead?" That's the race. Oracle just fired the starting gun.