A pattern keeps showing up in research and in the field, and it validates what I am hearing from customers of all industries, and in all regions worldwide. A recent report surveying data leaders across Australia and Singapore found that 97% of organizations deliver data products, yet only 25% do so through a structured, repeatable program.1 In conversations with data leaders across North America and Europe, I hear the same story: the gap between data ambition and deployment is wider than most boards realize.
The remaining 75% from the study operate ad hoc, in a fragmented mix, and for most, up to half their data team’s capacity goes toward one-off work that will never be reused or shared, capacity that is unavailable for the AI initiatives that actually compound in value.
New data and analytic projects assume that if the data exists it’s aligned and accurate. Typically, the conversation focuses on features and dashboards, not the question that actually determines success: What does this solution require of my data before I can trust it? And once it is trusted, how do I manage to keep it aligned to that new deliverable?
These are the five requirements that separate organizations that scale AI from those still rebuilding the same dataset every quarter.
1- Context and Governance Built into Creation, Not Inspected Afterward
The instinct to move fast on AI and govern later almost always backfires. As your project begins, capture and collect business information and incorporate enterprise data standards in a logical model. This doesn’t have to be a month-long project, in today’s data modeling solutions that use AI it’s fast and efficient. Combine your logical representation with metadata and you now have an inventory of what data to pull. The logical information combined with the physical metadata and governance (terms, policies, business rules) form an ontology which produces context in which your LLM’s, AI models and agents will demand.
Even organizations with formal, structured programs struggle here: design and modeling delays remain a leading bottleneck, cited by 45% of structured operators. The same gap shows up at the broader AI level: a separate IDC survey found that 78% of organizations claim complete trust in AI, yet only 40% have invested in the governance and safeguards to make that trust warranted.2
2- Automated Detection of What Already Exists
Differing requirements, where teams unknowingly build parallel versions of the same data product or asset, is the top cause of duplication among ad-hoc operators, at 74%. A platform that cannot flag a similar existing product before someone builds it is solving the wrong problem. The fix is active detection that surfaces reusable assets before recreation.
Collaborating on data products increases the quality of the data product. Social collaboration such as “ratings, contribution threads, extending and upgrading” data products keep them active as you re-use. Shop, sharing, and comparing in a familiar Amazon like forum keeps users coming back.
3- Explainable, Not Arbitrary, Trust Signals
Trust or quality issues, where teams would rather rebuild than reuse someone else’s data, rank as the top cause of duplicated work among structured operators, cited by 71%. A vague “approved” label does not solve that; a transparent, explainable scoring framework does, letting an analyst judge fit without tracking down the original owner. IDC’s research on AI trust found the same misalignment at the enterprise level: 47% of North American organizations show a gap between how much they trust their AI and how trustworthy it actually is.2
4- Discoverability Across Organizational Silos
A data product nobody can find is functionally a one-off, no matter how well it was built. Lack of awareness of existing data assets was cited by roughly half of respondents across every delivery model, a pattern Omdia independently confirmed in North America at 47%.3 That tells me discoverability is not a maturity problem organizations grow out of on their own. The requirement is a shared discovery layer that makes existing data products visible across teams, not just within the team that built them.
5- AI-Blended Capabilities That Remove Bottlenecks
The most common reason organizations resist structured delivery is that it creates a bottleneck and that effort overrides the benefit. Governance is seen as a tax on speed. However, explainability and compliance remain non-negotiable. Data trust can only be created through transparency and business alignment. That trade-off of speed versus trust should not be necessary, you need both.
AI is merging the swim lanes of our technical and business roles and responsibilities. Software solutions need to follow suit, which is why you see more data management platforms arising. Gone are the days of fragmented, point product solutions for capabilities like data lineage, business glossaries, data quality, MDM and yes, catalogs. These are now the base commodity level of a data management platform.
Today vendors are using AI to blend capabilities together instead of one group doing the modeling, one group cataloging, and another doing data governance. These capabilities can be blended just as the roles between technical and business users begin to blur. If speed and trust are still presented as a trade-off in 2026, that is a sign of an immature solution, not an inherent limitation.
Where This Leaves Any Evaluation
These five requirements are not abstract to me. They describe the architecture I work with every day, where automated modeling, trust scoring, context generation and blending capabilities with AI work together rather than as separate purchases, the design principle behind the Quest Trusted Data Management Platform, a single, converged solution for data product delivery.
Before you take this framework to a vendor, take it to your own team first. Ask which of the five you are already failing internally. That answer tells you whether you need a platform change or a process change.
Then make vendors show you, not tell you: create a data product live and watch whether lineage appears automatically, search for an existing asset before building a duplicate, and ask an analyst, not a sales engineer, whether a trust score makes sense without a phone call.
If your team is still ad hoc, start with detection and discoverability, they compound fastest. If you already run a structured program, governance and explainability matter more.
Stewart Bond, Vice President of IDC’s Data Intelligence and Integration Software Service, frames the business case directly: “By providing real-time, governed, contextualized data products, converged data management platforms can contribute to organizations enjoying higher levels of innovation, faster time to value, and improvements in financial business KPIs.”
Do not start with a vendor’s feature list. Start with what you cannot get answered honestly in the room.
—————-
Sources:
- Corinium Intelligence, “Transforming Data Delivery for AI Readiness, Australia and Singapore,” June 2026
- IDC eBook, sponsored by SAS, “Data and AI Impact Report: The Trust Imperative, with insights provided by IDC,” September 2025
- Omdia, “Scaling AI and Business Value Through Automated Data Products,” April 2026
