Every AI demo works. Almost none of them reach production

The pattern repeats across enterprises: a generative AI pilot impresses in the boardroom, then stalls the moment it meets a security review, a compliance officer, or a quarterly cloud bill. The problem is rarely the model. It is that most AI tooling was never designed for organizations that answer to auditors.

Three questions stop most enterprise AI initiatives:

01. Who owns what the AI produced?

When AI generates thousands of lines of opaque code, nobody can review it, version it, or take responsibility for it. It runs until it fails, and then no one can explain why. For a system that runs your operations, that is not a risk any IT director can accept.

02. Where does the data go?

Most AI features quietly assume your business data can travel to an external provider's cloud. For banks, insurers, healthcare, and the public sector, that assumption ends the conversation before it starts.

03. What does it cost, and who is accountable when it acts?

AI usage costs arrive as an unpredictable line item nobody budgeted. And when an AI assistant can act on live systems, "the model decided" is not an answer that survives an audit.

Oracle APEX 26.1, the largest release in the platform's history, was designed around these three questions.

AI that generates intent, not opaque code

Released on May 14, 2026, with release notes running over 90 pages, Oracle APEX 26.1 marks a clear direction: Oracle now positions APEX as the AI Application Generator. But the design principle behind it matters more than the branding. In APEX 26.1, AI does not produce piles of arbitrary source code or run free against your database. It generates application intent: structured, human-readable metadata that your team can review, version, validate, and govern before anything reaches production.

Oracle summarizes the philosophy in one sentence: if you generate it, you own it. AI proposes, the platform validates, and your people stay in control. Reports adjusted with natural language never execute AI-generated SQL. AI agents act only through tools you have explicitly approved. AI-assisted changes stay visible, inspectable, and reversible.

And because APEX runs inside the Oracle Database you already license, this entire AI layer arrives without a new platform to buy, with your existing security model intact, and with the option to keep models entirely within your own infrastructure. That combination, capability plus governance plus economics, is why we believe 26.1 is the release that makes enterprise AI deployable, not just demonstrable.

Case study

BRANCH

Insurance / HealthTech

COUNTRY

Greece

SYSTEM

Insurance platform (SaaS)

Munich Re HealthTech: 4 months to modern SaaS

A legacy insurance system rebuilt on Oracle APEX in four months. Critical analytics accelerated from days to minutes, and 15-day Excel-based processes now run as 20-minute APEX workflows. Read how the platform cut maintenance and licensing costs while staying secure and scalable.

Four capabilities that change what you can build

Oracle APEX 26.1 introduces dozens of improvements across workflow, security, translations, maps, and Universal Theme, but four capabilities define the release. Here is what each one makes possible in business terms.

Capability #1

APEXlang: your applications become reviewable source code

Consider a pull request for a business application. A developer, or an AI coding assistant, proposes a change. Your team sees exactly what will change, line by line, reviews it, and approves or rejects it before anything touches production.

APEXlang makes this the standard way of working. It is an open, declarative specification language that represents an entire APEX application as human-readable .apx text files. Applications can be exported, stored in Git, diffed, validated with Oracle SQLcl, edited in VS Code, and imported back. AI assistants get a supported, governed format they can generate and modify, and everything they produce goes through the same review as human work.

For organizations that want to enforce it, the new Application Lock option goes further: it prevents changes through the App Builder entirely, making your Git repository the single, enforced source of truth for the application.

What it means for your business: real source control, code review, and CI/CD for low-code applications, and a safe foundation for AI-assisted development at team scale.

Capability #2

AI Agents: assistants that act, within limits you define

A user types: "register a complaint for this order and notify the account manager." The agent understands the request, calls the approved tools that create the complaint record and send the notification, and reports back. If a tool requires confirmation, the user explicitly approves the action before it executes. If they decline, the action does not execute.

AI Agents in APEX 26.1 reason over user requests and take action through AI Tools: specific, developer-approved capabilities that can retrieve data, run server-side PL/SQL, or request a client-side interaction. The agent handles the conversation; the tools define what it is allowed to do. APEX manages the entire execution flow, and agent actions are designed to stay within the scope and security rules of the application.

What it means for your business: conversational interfaces on top of real business operations, with the boundaries, approvals, and audit trail that let them pass an internal review.

Capability #3

AI Interactive Reports: analytics in plain language

An operations manager opens a report and types: "European customers grouped by industry, highlight those with more than 50 open cases." The report rebuilds itself. Every change the AI applied appears as a visible, editable chip, so the user sees exactly what happened and can adjust or remove any step manually.

This is the detail that matters: APEX translates natural language into native, declarative Interactive Report settings, filters, sorting, control breaks, aggregations, pivots, and charts. It never executes AI-generated SQL, and it operates entirely within the report's existing data access rules.

What it means for your business: the report backlog that has been sitting in your IT queue for years starts to shrink, because business users can answer their own questions without waiting for a developer.

Capability #4

Data Reporter: self-service reporting with governance built in

Not every question fits an existing report. Data Reporter is a new reporting experience that lets business users create and manage their own reports without any APEX development skills. Administrators define trusted datasets and control who can access them; editors build and publish reports through a streamlined self-service interface.

The governance model is the point: the business gets autonomy, IT keeps control of what data is exposed and to whom.

What it means for your business: a genuine self-service analytics layer on your Oracle data, without buying and licensing a separate BI platform for casual reporting needs.

Also in Oracle APEX 26.1:

Workflow engine upgrades:
parallel flows, an AI text generation step inside processes, smarter task delegation, and new APIs for automation at scale.

Structured AI output:
AI responses validated against a JSON Schema, so AI becomes a reliable component in integrations, not just a chat window.

Your choice of AI model:
connect the leading commercial models or run private, locally hosted models so data never leaves your infrastructure.

Universal Theme 26.1:
the new Iris look, improved accessibility, and Font APEX 2.5 for a modern, compliant user interface.

Built for organizations that answer to auditors

Most AI platforms treat governance as a limitation to work around. Oracle APEX 26.1 treats it as the core design principle, and that maps directly onto what European regulators now expect from AI in business systems. Four mechanisms make this concrete:

DORA and regulated industry compliance 2

Explicitly scoped AI

Agents act only through approved AI Tools, each exposing one defined capability. There is no general-purpose access to your database or your application. What the AI is allowed to do is a decision your team makes, tool by tool.

Human oversight, at two levels

Individual agent actions can require explicit user approval before they execute. Entire business processes can route through Human Tasks, APEX's native mechanism for approvals and delegation. Human-in-the-loop is not a policy document; it is a platform feature.

A trail you can show an auditor

AI-assisted report changes appear as visible, editable steps. APEXlang files record what changed in every application version. Combined with database-level auditing, you can reconstruct what the AI did, when, and under whose approval.

Your data, your infrastructure

APEX connects to the AI provider of your choice, including private and locally hosted models running inside your own environment. For regulated industries, this is the difference between an AI initiative that starts and one that ends at the first security review.

What 26.1 does not do:
We work with Oracle APEX every day, so we will be direct about the boundaries. APEX 26.1 does not include a built-in AI model; you choose and connect one, and the quality of AI features depends on that choice. AI-generated applications still need engineering: review, refactoring, security hardening, and testing. And "if you generate it, you own it" applies in both directions: the platform gives you governance mechanisms, but using them well is an organizational discipline, not a checkbox. That discipline is precisely what our services are built around.

Where are you today? Start from your situation

In a regulated industry

Banking, insurance, telecom, healthcare, public sector: for you, AI adoption is inseparable from compliance obligations, including the EU AI Act. The architecture has to be right before the first feature ships.

How Pretius helps: from a two-day scan to running your AI in production

Our Oracle APEX 26.1 services are built in three layers. Fixed-price assessments give you clarity in days. Transformation projects deliver the upgrade, the AI capabilities, and the new way of working. Ongoing services keep it all running, governed, and cost-controlled.

Layer 1

Start here: fixed-price assessments

Low-commitment, high-clarity engagements. Each one ends with a concrete deliverable and a prioritized roadmap you own, whether or not you continue with us.

APEX 26.1 Readiness Scan

Fixed price | 1–2 days

An automated and expert review of your APEX estate against 26.1 requirements: platform and ORDS versions, database prerequisites, application and plug-in inventory, and identification of breaking changes, including custom JavaScript patterns and deprecated components that 26.1 affects.

You get:
a readiness report with a per-application risk rating and a recommended upgrade sequence.

Upgrade Assessment & Migration Roadmap

Fixed price | 3–5 days

A full pre-migration audit: applications, customizations, plug-ins, integrations, environments, and themes. For instances below 18.1, we resolve the key decision between a two-step upgrade and a fresh installation with application migration, two paths with very different cost profiles.

You get:
a complete migration roadmap with timeline, effort estimate, regression testing plan, and rollback strategy.

AI Cost & Governance Audit

Fixed price | 2–3 days

AI features in APEX consume tokens, and organizations that switch AI on without a cost architecture discover the problem in their second or third monthly bill. We measure actual and projected usage per feature and per user, design limits and caching strategies, and recommend model routing so each feature uses the most economical model that does the job.

You get:
a cost model, a governance framework for AI usage, and a monitoring dashboard design.

Layer 2

Transformation projects

Scoped, senior-led projects that take you from roadmap to running systems.

Upgrade & Migration Delivery

Execution of the migration defined in your roadmap: environment preparation, the upgrade or fresh installation itself, regression testing, a Universal Theme refresh to the new Iris style, stabilization, and hypercare. From single-instance upgrades to multi-year modernization programs covering dozens of legacy applications.

APEXlang DevOps Transformation

We bring real software engineering discipline to your APEX practice: Git repositories with .apx files as the source of truth, branching standards, SQLcl validation, CI/CD pipelines, code review for application changes, and optionally Application Lock to enforce it all. Available as a greenfield standard for new work or a brownfield program that brings your existing applications into the pipeline.

Secure AI Agents Implementation

Architecture and delivery of AI agents for organizations where "it usually works" is not acceptable: a controlled agent loop, whitelisted operations behind an AI-to-PL/SQL gateway, prompt injection defenses, sensitive data separation, full audit trail, human approval of agent actions through Human Tasks, and adversarial testing before go-live. Designed for banking, insurance, telecom, and public sector requirements.

Self-Service BI: AI Reports & Data Reporter

AI Reports & Data Reporter Deployment of natural language reporting on your existing Interactive Reports and a governed Data Reporter rollout: trusted datasets, roles, access governance, and the business context that makes AI answers accurate. Cost optimization is designed in from the start, not patched in after the first invoice.

GenAI Discovery Accelerator

We use 26.1's generative capabilities where they create the most leverage: the analysis phase. Requirements workshops where application skeletons are generated live, a prompt playbook for your analysts, and a refactoring checklist that separates a useful prototype from production-ready software. We shorten discovery and protect you from AI-generated technical debt at the same time.

Workflow & Human Tasks Automation

Business process automation on the 26.1 workflow engine: parallel flows, AI text generation steps inside processes, intelligent delegation, and multi-tenant process architectures. For Oracle-centric organizations, an honest alternative to licensing a separate BPM platform, and we will tell you when a dedicated BPM engine is genuinely the better fit.

Layer 3

Ongoing services

APEX estates are not projects; they are living systems. We stay accountable after go-live.

APEX Managed Services

SLA-based maintenance of your applications and instances: patch set bundle management, monitoring, incident response, and continuous small-scale development. Your APEX estate stays current, secure, and supported, release after release.

Trusted Advisor Retainer

Ongoing architectural oversight of your AI and DevOps practice in APEX: quarterly reviews, change opinions, and policy updates as new APEX versions and regulatory guidance arrive. Direct access to the people who work with the platform at Oracle ACE level.

Not sure where 26.1 fits in your roadmap?

Book a no-obligation discovery call. In 45 minutes, we will map your current APEX estate, identify what 26.1 changes for your applications and your AI plans, and tell you which starting point makes sense, based on your organization, not ours.

Book a discovery call

Why Pretius

We are recognized by Oracle itself

Five members of our board hold Oracle ACE titles, including Matt Mulvaney, one of only 17 Oracle ACE Directors worldwide specializing in APEX. ACE status means a direct dialogue with Oracle product management: we build on roadmap knowledge, not release notes.

We wrote the playbook on 26.1

Our team publishes reference technical material on APEX 26.1: AI cost management, secure agent architecture, upgrade strategy, and workflow automation. When you work with Pretius, the people who wrote the reference articles are the people on your project.

We run mission-critical APEX every day

The platforms we build and maintain for clients like Munich Re HealthTech and Sweco operate around the clock. We know the difference between a feature that impresses in a demo and one that survives production, and we apply that standard to every AI capability we ship.

We tell you what not to do

APEX 26.1 ships without a built-in AI model; quality depends on the choices you make, and some AI use cases are simply not worth building. Our recommendations follow your data, your team, and your strategy, including the moments when the honest answer is "not this, not yet."

Cost discipline from day one

We treat AI spending as an architecture decision, not an afterthought. Limits, caching, and model routing are designed before launch, and our AI FinOps practice keeps costs visible and controlled afterward.

Scale and long-term partnership

300+ specialists, more than 70 focused on Oracle APEX. Offices in Warsaw and London. 93% of clients continue working with us after the first project. We are built for the scale that enterprise APEX estates require, and we are here for what comes after go-live, too.

Our process

01 Readiness Scan

We start with facts, not opinions: a fixed-price scan of your APEX estate against 26.1 requirements, breaking changes, and AI readiness. Days, not months.

02 Roadmap & recommendation

Based on the scan, we present a clear recommendation: upgrade path, AI adoption sequence, and governance model, with tradeoffs and effort estimates on the table. You decide with full information.

03 Pilot on a real use case

We prove the approach on one meaningful scenario, a report backlog, an agent workflow, or an APEXlang pipeline for a single team, with governance and cost controls active from the first day, not retrofitted later.

04 Phased delivery

Upgrades and AI capabilities roll out application by application, in two-week sprints with regular stakeholder demos. Production stability is a constraint, not a hope.

05 Enablement & run

Your team is trained on the new toolchain, or we keep running it with you: managed services, AI FinOps, and architectural oversight under one accountable partner.

Flexible project approaches

01 Co-creation

Our preferred model for complex, business-critical work. Our engineers work alongside your in-house team from day one. Knowledge transfers continuously. When the project ends, your team understands the new toolchain deeply enough to own it, not just use it. This model was central to the work we carried out for a big international bank, where Pretius developers and the bank's internal team formed a single unit.

03

02 Dedicated development teams

You have a clear scope and a clear destination. We provide a cross-functional team of developers, analysts, QA, and DevOps that operates as an extension of your organization. You maintain strategic control; we bring the execution capacity.

02

03 Full project delivery

You hand us the project, we deliver the outcome. Defined scope, defined timeline, defined budget. Minimal internal overhead for your team. Optional maintenance SLA after go-live.

04 Team augmentation

You have an internal team running the adoption but need specific expertise: APEX architecture, APEXlang and CI/CD, AI integration, or database performance. We add specialists to your team for as long as you need them.

What our clients say

Bilal Ramadan

Chief Executive Officer at Munich Re HealthTech

The solution is secure, scalable, and compatible with most technologies used by Munich Re HealthTech’s clients. It’s also much cheaper to maintain which was an important success criteria factor. Application licensing costs were largely reduced.

Radosław Olkiewicz

CEO at Profitowi

Thanks to the new client needs analysis system, our sales process is much easier and standardized. It helps our Experts better tailor insurance or financial solutions to the needs of the client. This results in increased sales to new clients and smooth upselling to existing ones.

Przemysław Wilary

IT Project Manager at Play

We were impressed with Pretius' experience in the telecom industry. They proposed a few unique solutions that enabled our employees to configure individual offers faster than ever before.

Marek Samujłło

CIO of Emitel

What I like the most, as the CIO, is that the people at Pretius care not only about technology itself but rather about how this technology can help your company achieve business goals.

Zuzanna Świętorzecka

Head of Bancassurance at BNP Paribas

The communication throughout the project was very good as Pretius provided us with a Project Manager who always addressed our needs. The most impressive thing was how quickly the team developed a fully working, market-ready software solution for us.

Konrad Gałecki

Digital Manager at Philip Morris Distribution

Usually, you can either get a high-quality software solution or have it developed fast. However, thanks to choosing the right technology, the Pretius team was able to provide us with a fully working platform in a relatively short time.

Artur Męcina

IT Manager at T-Mobile Poland

The most important thing about Pretius is that they care to understand your business goals. And, having a bunch of enterprise-grade and SME-grade projects under their belt, they’re very good at it.

Renata Kania

Senior Security Awareness Specialist at T-Mobile

Pretius has quite a few huge, enterprise-grade projects in their portfolio. I believe the reason for this is that they go the extra mile to understand the partner’s business goals and do what they can to provide a software solution that meets these goals.

Piotr Blaszczyk

Executive Director of EBS

Pretius has a lot of experience in dealing with big, enterprise-grade companies like ours. They’ve proved they have a keen understanding of our needs and business goals.

Adam Kasperowicz

CIO at VAN group

The new e-VAN system has become an innovative tool and work environment. The development team allows us to use the potential of the latest trends and technological solutions: web & mobile development, responsiveness, integrations, low-code, and cloud computing.

Graem Lourens

Co-Founder & CTO at Lourens Systems

They’ve garnered our confidence and loyalty through their technical skills and the deep involvement.

Marijn van den Berg

Team Manager of Obsurv Software at Sweco

Thanks to our collaboration with Pretius, we've made it easier for our clients to utilize Obsurv. Customizing the application modules to their specific needs is now much quicker and more efficient. For us, this means shorter customer onboarding times and lower costs.

Gordian Ratajczak

Product Owner at Neuroflash

They effectively assess a project and allocate the matching resources, which gives confidence that you are in good hands. Issues and tasks are handled proactively and communication stands on a firm ground of mutual understanding and goals. You are not hiring an external resource, but you are gaining a team member.

Esteban Michalowski

CTO at AltaPay

I’ve always found interactions with Pretius (previous IN Team) to be both streamlined, light, and effective. Pretius always seeks ways to improve, and have a process for ensuring the collaboration is going according to expectations.

Richard Coombes

Head of Engineering at Equals Money

They allowed us to deliver against our 20/21 roadmaps, launching new products to market and increasing company revenue by over 70%.

Every organization will adopt AI in its applications. Few will do it in a way that survives an audit.

Oracle APEX 26.1 gives you the platform mechanics: governed agents, reviewable AI output, human oversight, and reporting in plain language, all inside the database you already run. What it cannot give you is the adoption strategy, the cost architecture, and the engineering discipline. That is what we do.

Book a discovery call

Or use the form below

FAQ

Oracle APEX 26.1 is the May 2026 release of Oracle APEX, the low-code platform included with the Oracle Database, and the largest release in the platform's history. Its headline capabilities are APEXlang (a human-readable specification language for APEX applications), AI Agents (assistants that act through developer-approved tools), AI Interactive Reports (natural language data exploration), and Data Reporter (governed self-service reporting). Oracle positions APEX 26.1 as its AI Application Generator.

No. APEXlang is an additional representation of your applications, not a replacement for the App Builder. Any existing application can be exported to APEXlang, and teams can keep building visually while using .apx files for version control, review, and AI-assisted changes. Organizations that want stricter governance can adopt it gradually, one team or one application at a time.

Not on its own. AI Interactive Reports never execute AI-generated SQL; natural language requests are translated into declarative report settings within existing data access rules. AI Agents act only through explicitly approved tools, and individual actions can require user confirmation before they execute. What AI is allowed to do is defined by your team, tool by tool.

APEX 26.1 has no built-in AI model; you connect the provider of your choice. That includes leading commercial models as well as private and locally hosted models running entirely within your own infrastructure, which is the configuration we typically recommend for regulated industries.

APEX 26.1 requires Oracle Database 19c (with Release Update 19.18 or newer) or Oracle Database 26ai, plus ORDS 26.1.1 or newer. Direct upgrades are supported from APEX 18.1 onward; older instances need a two-step upgrade or a fresh installation with application migration. Our Readiness Scan verifies all of this against your actual estate in one to two days.

Oracle APEX is a fully supported, no-cost feature of the Oracle Database. If you hold a supported Oracle Database license, APEX 26.1 and all of its AI capabilities are included; the only AI-related costs are the model usage fees of the provider you connect, which is what our AI cost governance services are designed to keep under control.

Yes, and 26.1 is arguably the first low-code release designed with that reality in mind. Explicitly scoped tools, human approval of agent actions, Human Tasks for process-level oversight, and detailed audit trails map directly onto regulatory expectations, including the obligations that apply to organizations deploying AI under the EU AI Act. Our AI Security & Compliance Review assesses your specific use cases against those requirements.

Oracle APEX 26.1: AI your organization can actually deploy

Every platform now promises AI. APEX 26.1 is built for the harder part: AI that your security team, your auditors, and your CFO can all sign off on. We help you put it to work, from a two-day readiness scan to secure AI agents running in production.

Every AI demo works. Almost none of them reach production

Three questions stop most enterprise AI initiatives:

01. Who owns what the AI produced?

02. Where does the data go?

03. What does it cost, and who is accountable when it acts?

AI that generates intent, not opaque code

Case study

BRANCH

Insurance / HealthTech

COUNTRY

Greece

SYSTEM

Insurance platform (SaaS)

Munich Re HealthTech: 4 months to modern SaaS

Four capabilities that change what you can build

Oracle APEX 26.1 introduces dozens of improvements across workflow, security, translations, maps, and Universal Theme, but four capabilities define the release. Here is what each one makes possible in business terms.

Capability #1

APEXlang: your applications become reviewable source code

Capability #2

AI Agents: assistants that act, within limits you define

Capability #3

AI Interactive Reports: analytics in plain language

Capability #4

Data Reporter: self-service reporting with governance built in

Also in Oracle APEX 26.1:

Built for organizations that answer to auditors

Explicitly scoped AI

Human oversight, at two levels

A trail you can show an auditor

Your data, your infrastructure

Where are you today? Start from your situation

Running APEX older than 18.1, or still on Oracle Forms

On APEX 23.2 or older supported versions

On APEX 24.x, planning your adoption

Under pressure to "have AI" now

In a regulated industry

How Pretius helps: from a two-day scan to running your AI in production

Our Oracle APEX 26.1 services are built in three layers. Fixed-price assessments give you clarity in days. Transformation projects deliver the upgrade, the AI capabilities, and the new way of working. Ongoing services keep it all running, governed, and cost-controlled.

Layer 1

Start here: fixed-price assessments

APEX 26.1 Readiness Scan

Upgrade Assessment & Migration Roadmap

AI Cost & Governance Audit

Layer 2

Transformation projects

Upgrade & Migration Delivery

APEXlang DevOps Transformation

Secure AI Agents Implementation

Self-Service BI: AI Reports & Data Reporter

GenAI Discovery Accelerator

Workflow & Human Tasks Automation

Layer 3

Ongoing services

APEX Managed Services

Trusted Advisor Retainer

We are a top-rated software development company for a reason

We keep your data safe: ISO certified

Not sure where 26.1 fits in your roadmap?

Book a no-obligation discovery call. In 45 minutes, we will map your current APEX estate, identify what 26.1 changes for your applications and your AI plans, and tell you which starting point makes sense, based on your organization, not ours.

Why Pretius

We are recognized by Oracle itself

We wrote the playbook on 26.1

We run mission-critical APEX every day

We tell you what not to do

Cost discipline from day one

Scale and long-term partnership

Our process

01

Readiness Scan

02

Roadmap & recommendation

03

Pilot on a real use case

04

Phased delivery

05

Enablement & run

Flexible project approaches