Australia’s payments infrastructure isn’t exactly simple. Australian Payments Plus (AP+) — the organisation behind NPP Australia, BPAY Group, and eftpos — manages some of the most critical financial plumbing in the country. Millions of transactions, dozens of regulatory obligations, legacy technical debt layered over newer real-time rails. So when AP+ went public with details of how it’s been using ChatGPT Enterprise and OpenAI Codex to work faster, it was worth paying close attention. This isn’t a startup running experiments. This is a tier-one payments infrastructure operator making production bets on AI — and the specifics tell us a lot about where enterprise AI adoption is actually heading in 2026.
Why AP+ Needed AI Help in the First Place
Payments organisations sit at the intersection of extreme technical complexity and extreme regulatory scrutiny. AP+ inherited three distinct legacy businesses when it merged in 2022, which means three sets of engineering cultures, codebases, documentation standards, and compliance workflows. Getting anything done — a new feature, a standards update, a regulatory submission — requires coordinating across all of that at once.
That kind of complexity creates real drag. Engineers spend time reading through documentation that hasn’t been touched in years. Compliance teams manually cross-reference regulatory requirements against product specs. Developers write boilerplate code that could be generated in seconds. None of this is glamorous work, and none of it directly moves the needle on what AP+ is actually trying to build.
The problem isn’t unique to AP+. Any large financial institution running on aged infrastructure faces this. But what’s interesting about AP+’s approach is that they didn’t try to boil the ocean. They identified specific, high-friction workflows and targeted them. That’s a much more honest and useful story than the typical “we’re transforming everything with AI” press release.
What AP+ Actually Built with ChatGPT Enterprise and Codex
According to OpenAI’s case study on Australian Payments Plus, the deployment spans both knowledge work and software development — with ChatGPT Enterprise handling the former and Codex increasingly taking on the latter.
Accelerating Payments Documentation and Compliance Work
One of the biggest time sinks in any payments organisation is documentation: reading it, writing it, reconciling it against standards, and translating dense regulatory language into something engineers can actually act on. AP+ teams have been using ChatGPT Enterprise to compress this work significantly.
Think about what that means in practice. An engineer trying to understand how a particular payment flow maps to ISO 20022 messaging standards doesn’t need to spend an afternoon digging through PDFs. A compliance analyst drafting a response to a consultation paper can use ChatGPT to draft the first version and then apply their own expertise to refine it. The model handles the grunt work; humans handle the judgment calls.
This matters because human judgment stays central — which is the line AP+ keeps emphasising, and it’s worth taking seriously. Payments isn’t a space where you want AI making autonomous decisions about what gets approved or rejected. The goal is freeing up expert time, not replacing expert judgment.
Using Codex to Speed Up Development Cycles
On the engineering side, AP+ has deployed OpenAI Codex — OpenAI’s AI software engineering tool — to help developers move faster through the kinds of coding tasks that slow down sprints without adding strategic value. That includes writing unit tests, generating boilerplate, and helping developers get up to speed on unfamiliar parts of the codebase.
Here’s the thing: payments code is notoriously conservative. You don’t refactor a transaction processing service on a whim. But that conservatism means a huge amount of time gets spent on the surrounding scaffolding — the tests, the documentation, the helper functions — rather than the core logic. If Codex can take 60-70% of that load, developers can redirect their attention to the parts that actually require deep domain knowledge.
Key areas where AP+ reported gains include:
- Code generation and boilerplate: Reducing time spent on repetitive coding tasks that follow predictable patterns
- Test coverage: Using Codex to generate unit and integration tests, improving quality without proportionally increasing developer time
- Documentation generation: Automatically producing inline documentation and technical specs from existing code
- Onboarding acceleration: Helping new engineers understand legacy codebases faster by querying the AI about how specific systems work
- Standards translation: Converting complex payments standards language into developer-friendly implementation guidance
The Enterprise Guardrails That Made This Viable
AP+ didn’t deploy consumer ChatGPT. They went with ChatGPT Enterprise, which matters for a few reasons. Enterprise comes with data privacy guarantees — conversations aren’t used to train OpenAI’s models, and data doesn’t leave the organisation’s controlled environment. For a company handling sensitive payments infrastructure information, that’s not optional.
It also comes with admin controls, usage analytics, and the ability to set organisation-wide policies around how the tool gets used. In a regulated industry, being able to audit AI usage and demonstrate appropriate controls to regulators is a real requirement, not a nice-to-have.
What This Actually Means for Fintech AI Adoption
AP+’s deployment is interesting not because it’s dramatically different from what other enterprises are doing, but because it’s so representative of where serious AI adoption is landing right now. The headline-grabbing stuff — fully autonomous AI agents, AI making credit decisions, AI replacing entire teams — isn’t what’s happening at sophisticated organisations. What’s happening is more like this: targeted deployment in high-friction, lower-risk workflows, with humans remaining in the loop on anything that touches compliance, customer outcomes, or critical system changes.
That’s actually a healthy pattern. It builds organisational confidence in AI tools before extending them into higher-stakes territory. It generates measurable ROI that justifies broader rollout. And it forces teams to develop the internal literacy to use these tools well, rather than just deploying them and hoping for the best.
I wouldn’t be surprised if AP+’s next phase involves more agentic workflows — having AI systems that can autonomously draft a compliance response, run it through an internal review workflow, and flag it for a human to approve, rather than requiring a human to initiate each step. Managed agents with background task capabilities are already showing up in enterprise tools, and the demand is clearly there.
The competitive dynamics are worth watching too. OpenAI isn’t the only player chasing enterprise fintech contracts. Microsoft Copilot (built on GPT-4o) is already embedded in many financial institutions through existing Office 365 relationships. Google’s Gemini for Workspace is making its own enterprise push. The question for OpenAI is whether ChatGPT Enterprise’s data privacy stance, model quality, and Codex integration can hold the line against Microsoft’s distribution advantage in corporate environments. AP+ is a meaningful win, but the fight for enterprise AI contracts in financial services is far from settled.
It’s also worth noting what AP+ didn’t say. There’s no claim of massive headcount reduction. No suggestion that AI is replacing payments expertise. The framing throughout is about speed and quality — doing the same work faster and with fewer errors, while keeping experienced humans accountable for the outputs. That’s a more defensible and honest position than the productivity miracle narratives that have defined a lot of enterprise AI marketing over the past two years. For more on how those adoption numbers are actually playing out, our breakdown of ChatGPT’s enterprise adoption trajectory is worth a read.
Key Takeaways for Enterprises Watching This Deployment
- Start with documentation and knowledge work: AP+ found immediate ROI in tasks that involve reading, synthesising, and producing structured text — before touching core system logic
- Pair ChatGPT Enterprise with Codex for end-to-end coverage: Knowledge work and software development have different AI needs; using specialised tools for each yields better results than a one-size-fits-all approach
- Data privacy controls are non-negotiable in regulated industries: Enterprise licensing isn’t just a pricing tier — it’s the difference between a tool you can actually use and one that creates regulatory exposure
- Human judgment stays in the loop: AP+ consistently emphasised this, and it’s the right framing for any deployment in financial services
- Measure before expanding: Targeted deployments with measurable outcomes build the internal case for broader rollout far more effectively than big-bang transformations
Frequently Asked Questions
What is ChatGPT Enterprise, and how is it different from regular ChatGPT?
ChatGPT Enterprise is OpenAI’s business-tier product that includes enhanced data privacy (conversations aren’t used for training), admin controls, higher usage limits, and access to the most capable models. For organisations in regulated industries like financial services, these controls are essential for compliance. Pricing is typically negotiated directly with OpenAI and scales with seat count.
What is OpenAI Codex, and what does it do?
OpenAI Codex is an AI software engineering tool built on top of OpenAI’s models, designed to help developers write, review, test, and document code. It can operate as an AI coding agent that works through tasks semi-autonomously. AP+ used it primarily to accelerate boilerplate generation, test writing, and onboarding into complex legacy codebases — the kinds of tasks that consume developer time without requiring deep domain expertise.
Is AP+’s AI deployment typical of what other financial institutions are doing?
Broadly, yes. Most large financial institutions are deploying AI in knowledge work and developer tooling before extending to customer-facing or decision-making systems. The pattern of targeted deployment in lower-risk workflows, with humans retaining accountability for outputs, is common across the sector. What varies is the pace and sophistication of the rollout — AP+ appears to be ahead of many peers in terms of genuine operational integration rather than pilot-stage experimentation.
Does using AI for payments documentation create compliance or security risks?
It can, if done carelessly. AP+’s use of ChatGPT Enterprise specifically addresses the main risk: sensitive information isn’t transmitted to a third-party training pipeline. The organisation retains control over what gets queried and can audit AI usage. That said, any enterprise deploying AI in a regulated environment needs clear internal policies around what information staff can share with AI tools — something AP+ appears to have built into their rollout from the start. OpenAI’s Enterprise Privacy documentation outlines the contractual protections in detail.
AP+’s story isn’t the most dramatic AI deployment you’ll read about this year — and that’s precisely what makes it credible. Payments infrastructure moves slowly by design, and the organisations that successfully integrate AI into it will be the ones that respect that pace rather than fighting it. As more of Australia’s financial sector watches what AP+ is doing, the real question is whether other institutions will follow the same disciplined, workflow-by-workflow approach, or whether they’ll try to shortcut the hard organisational work that makes these deployments actually stick. Given the growing number of major enterprises committing to OpenAI at scale, the pressure to move fast is only going to intensify. How AP+ manages the next phase of this rollout — particularly if it moves toward more autonomous agentic workflows — will be worth watching closely.
