TL;DR: New APAC research from Quest and Corinium reveals a surprising paradox: even organizations with mature, structured data programs are rebuilding data assets instead of reusing them, not because the assets don’t exist, but because nobody trusts them. Closing that gap requires visible trust signals, auditable governance, and a data marketplace that makes confidence the default.
Just six months ago, I wrote about the organizations I’ve watched fail at AI, not because of bad models or insufficient compute, but because nobody trusted the data feeding them. I argued that building AI for AI, using artificial intelligence to manage, govern, and serve data at scale, was the path forward. That trust infrastructure was the missing piece.
New APAC research1 commissioned by Quest Software in partnership with Corinium Global Intelligence suggests the problem runs even deeper.
A survey of more than 150 Chief Data and Analytics Officers and senior data leaders across Australia and Singapore surfaced a finding that stopped me in my tracks: even organizations that have done the hard work of building structured data programs — namely ones which have repeatable workflows, standardized delivery, and governance in place — are still rebuilding data assets rather than reusing the ones they have.
Not because they lack the tools. Because they don’t trust what’s already on the shelf.
And when you dig into why, the answer gets more uncomfortable. The research points to a cluster of systemic issues sitting beneath the trust deficit: data silos that prevent visibility across the organization, poor communication between departments, and a simple lack of awareness that relevant data assets even exist elsewhere. It’s not that teams distrust what they can see. It’s that the conditions for trust, such as data classification, shared context, managed data pipelines and cross-functional visibility, aren’t in place to begin with.
The finding that reframes the problem
Here’s what the data shows. Only 25% of respondents operate a structured data products program with standardized, repeatable workflows. The remaining 75% are working ad hoc or through hybrid approaches. For those ad hoc operators, duplication is largely what you’d expect: different teams, different requirements, no shared infrastructure. That’s a maturity problem, and the solutions are reasonably well understood.
When you look at what’s happening inside the 25% who’ve made it to structured delivery , the leading cause of duplication isn’t siloed teams or differing requirements. It’s trust and data quality concerns. Their teams are actively choosing to rebuild data assets rather than reuse work created elsewhere in the organization, even when that work exists, is findable, and was built to be shared.
I believe in a flywheel effect: a well-functioning data marketplace drives usage, usage generates ratings and feedback, ratings build trust, and trust drives more usage. It’s a self-reinforcing cycle that makes reuse the default behavior over time.
This new research reveals that most organizations haven’t got that flywheel turning. And the ones who have it in some form of motion are discovering that it stalls, not at the start, but mid-rotation when people don’t believe the assets already on the shelf are good enough to use.
What the trust signals are still missing
A comprehensive trust model has the following non-negotiables: data quality metrics, data curation, source origination authority, usage and social proof, and governance and sensitivity status. Five components. Each one visible, each one verifiable.
The new research reveals a stubborn gap between organizations having these components and being able to evidence them.
The survey respondents were significantly more confident in their ability to meet compliance obligations than they are in their ability to produce clear, auditable proof that those obligations have been met. Between 11 and 30% of IT spend is already absorbed by compliance activity, yet the evidence gap persists. Organizations are spending heavily on compliance and still can’t demonstrate it.
This matters because it maps directly to the trust gap I have previously written about , when IDC research showed that while 78% of organizations claim to fully trust their AI, only 40% have invested to make their systems demonstrably trustworthy through governance, explainability, and ethical safeguards.2
The semantic layer that gives AI your organizational context, allows it to apply your business rules, and keeps it from confidently suggesting something absurd will only work if the trust signals underpinning it are complete and visible. Most organizations have the framework. What they’re missing is the evidence layer that makes the framework real to the people deciding whether to reuse an asset or rebuild it.
The production gap is a trust gap
The cost of this isn’t theoretical anymore. Forty-four percent of surveyed organizations have active AI or GenAI work underway that depends on foundational data preparation. A further 45% are experimenting but not yet in production. That means nearly 90% of the market is on the same road, and most of them are stalled close to the starting gate.
The bottleneck isn’t compute. It isn’t tooling. It’s whether someone can open a catalog, find an asset, read its lineage and trust score, understand who owns it and when it was last validated, and make a confident reuse decision in under five minutes. Right now, most can’t.
Can you?
When AI is used to manage, harmonize, govern, and serve data at the speed and scale modern AI models demand – the definition of AI for AI – could compress data product delivery from four to six months down to two to three days. That is a 98% reduction in duration with a nearly 60x acceleration, or productivity gain. Those numbers are real. I’ve seen it in production. But the organizations that can’t get there aren’t being held back by the technology. The capability exists. The confidence doesn’t.
Think about how you shop on Amazon. You search, you compare, you check the ratings, you read what other buyers said, and then you decide. That’s exactly the model a data marketplace should offer. The research confirms that without visible trust signals, people default to the same behavior every time: they build their own.
Three steps forward
This new research sharpens what each step actually requires even if the overall path hasn’t changed.
First, don’t just package data as products, make trust visible inside the package.
Quality scores, classification tags, lineage transparency, social queues, and usage history aren’t optional metadata. They’re the difference between an asset that gets reused and one that gets rebuilt. If a consumer can’t evaluate trust in under five minutes, you haven’t built a data product. You’ve built a data artifact.
Second, close the evidence gap, not just the compliance gap.
Meeting an obligation and being able to demonstrate it are not the same thing. Build the audit trail as you build the data product, not as a retrospective exercise before a regulator asks for it. The organizations losing ground on compliance confidence are the ones treating governance as something you layer on at the end.
Third, design for reuse confidence, not just reuse availability.
If the path of least resistance is still rebuilding, you have a catalog, not a marketplace. The flywheel only turns when people trust what’s on the shelf enough to pick it up. That means investing in the social collaboration, the ratings, the lineage transparency, and the cross-team visibility that make trust legible rather than assumed.
The trusted data needed to move AI from pilot to production already exists somewhere inside an organization. Think about that. The problem is that nobody trusts it enough to use it. That’s not a technology problem. It’s an infrastructure problem, and it’s one we know how to solve.
Sources:
- Corinium Intelligence, “Transforming Data Delivery for AI Readiness, Australia and Singapore,” June 2026
- IDC eBook, sponsored by SAS, “Data and AI Impact Report,” EUR153787025, October 2025
