MongoDB’s Thorsten Walther says the real AI drain is hiding somewhere else entirely, and it’s costing Australian businesses far more.
If you are an Australian business that’s invested in AI in the past few years, something should have become clear by now: the challenge isn’t really about how to adopt AI, that’s pretty easy these days. The hard part is to run it—securely and efficiently—at scale.
Experts agree: many Australian organisations are getting stuck in the “pilot trap”. There’s a lot of great experimentation done with AI, but when it comes to moving these ideas and proofs of concept into day-to-day operations, things get trickier.
Deloitte’s State of AI in the Enterprise 2026 report shows that Aussie businesses are increasingly lagging behind their global peers when it comes to realising AI transformation at scale, pointing out that the real challenge is shifting from pilots to production.
This is compounded by the fact that many organisations are looking at the economics around AI implementations the wrong way. Most businesses look at the tip of the AI costs iceberg, and try to optimise what they can there.
But the hidden costs under the surface—where data feeding AI models and agents lies—are much more important and urgent to address. That is the key to moving from pilots to production, and ultimately scale.
The data cost: what’s visible isn’t what’s the most damaging
Many organisations have become fixated on the tip of the AI costs iceberg: model pricing, token bills, GPU costs, software licences, cloud consumption, and the latest invoice from another AI vendor.
These costs are real, and usually easy to measure. But the larger drag often comes from what’s happening under the surface to enable AI. It’s not just about dollar costs, it’s about the costs driven by inefficiencies and poorly-fed AI models.
For many, this is where the friction begins: the AI model is ready, but the data foundation is not.
A proof of concept can run on curated data, a limited workflow, and a small team of experts. Production AI has to work with the business as it exists. And once AI starts operating parts of the business, stale data and brittle integrations stop being an inconvenience. They become operational risks.
Too many Australian organisations are stockpiling data for the sake of AI, but all data isn’t equal, and not all data is built for meaningful AI. Cleaning bad data, maintaining fragile integrations, and duplicating (or de-duplicating) data across systems bear a great cost to the business.
A recent IDC research commissioned by MongoDB found that more than half (58%) of Australian organisations’ existing architectures make it impossible to build new applications without extensive modernisation, with poor data quality consistently appearing as a key issue.
An agent reasoning over a stale customer record, last week’s inventory, or a price that has already changed won’t just return a wrong answer. It will take a wrong action, and leave an audit trail of it.
Beyond the data itself, the architecture behind the scenes is also a critical barrier to moving to AI production at scale. Models and agents are only as good as the data feeding them, but success also depends on how efficiently and fast they can reach it.
That’s why data architecture, not model choice, increasingly decides who succeeds with AI: a foundation that holds structured, semi-structured and unstructured data together, evolves as schemas change, scales as demand grows, and serves current operational data the moment a decision is made.
The case for modernisation (and creating three times more digital revenue!)
The stakes for modernisation are now critical. The real AI cost conversation should not be “how do we make the model cheaper?”, but rather “how do we reduce the rework required to make AI useful?”
For technical decision makers, this means looking beyond the visible AI stack. The issue is not only which model to use, which cloud service to select, or which AI assistant to trial.
High-quality, integrated data is the essential fuel that determines the accuracy and performance of an AI application, making modern data architecture a foundational element of any AI strategy. Simplifying the data architecture means AI can use trusted, current operational data without every new use case creating another copy, pipeline, or workaround.
Not only does that save money (and headaches), it creates actual value: the IDC research showed that Australian leaders overcoming technical debt to unlock AI drive three times more digital revenue than their peers!
Three moves to drive less architectural waste now
- First, make data quality and governance non-negotiable. AI systems need consistent, trusted operational data, not a patchwork of stale extracts and hand-curated datasets. If teams cannot explain where data came from, how current it is, and who is allowed to use it, they are not ready for production AI.
- Second, modernise the architectures that block change. Lift-and-shift migration often moves old constraints into a new environment. That creates the impression of progress without removing the rigidity underneath. Organisations should prioritise architectures that support structured, semi-structured, unstructured, and vector data without forcing developers to stitch together a new stack for every AI use case.
- Third, treat modernisation as an operating discipline. Leaders invest in skills, change management, cloud-ready platforms, and clear business objectives. They do not wait for a perfect transformation window. They build the muscle to keep modernising as markets, regulations, customer expectations, and AI capabilities change.
The Australian businesses able to produce secure and reliable AI at scale aren’t the ones that negotiated the cheapest LLM license at the tip of the iceberg. They are the ones that had the foresight to fix the data architecture hidden beneath the surface. The organisations willing to modernise their architecture and build a unified data foundation will be the ones moving from pilot to production, and steering their AI investments toward true commercial value.
Keep up to date with our stories on LinkedIn, Twitter, Facebook and Instagram.
