Strategy and governance aren't the finish line. The decision is.

Executive Summary

Atrium's recent article “Data Strategy vs. Data Governance: Why You Need Both” makes an important distinction. Data strategy establishes what an organization wants to achieve with data. Data governance establishes the ownership, definitions, quality, security, policies and accountability required to use that data with confidence. Atrium's argument is that organizations need both: strategy without governance can produce ambitious programs built on information people do not trust, while governance without strategy can become disconnected from the outcomes the business actually cares about.

I agree, but I think there is another step worth exploring. An organization can have a clear strategy, mature governance, modern platforms, catalogs, lineage and quality controls, and someone in the business can still be asking: can I get the trusted data I need to make this decision? That is the gap. And as trusted data increasingly moves beyond dashboards into applications, copilots and AI agents, there is another requirement: the governance and business context that made the information trustworthy cannot disappear on the journey. It has to remain connected at the point of use. That is where strategy, governance and the decision finally meet.

The Article We're Responding To

In Data Strategy vs. Data Governance: Why You Need Both, Atrium's Riya Agarwal separates two disciplines that are often treated as interchangeable. Data strategy is about direction: what the organization wants data to accomplish, which outcomes matter, which capabilities are required and where investment should be focused. Data governance is about creating the conditions under which that strategy can operate reliably: who owns the data, what it means, how its quality is maintained, who can access it, what policies apply and how accountability is established.

The distinction matters because organizations can invest heavily in one without solving the problems addressed by the other. Strong strategy without effective governance produces sophisticated platforms, pipelines, dashboards and AI initiatives sitting on top of inconsistent definitions and uncertain ownership; the technology works, yet different parts of the business still arrive at different versions of the truth. Governance without strategy creates the opposite failure. An organization can catalog thousands of assets, assign owners, document definitions and implement quality rules without being clear about which business outcomes all of that work is meant to improve. The data may be governed, but governed for what?

Atrium makes both failure modes tangible. On one side is the organization pursuing a 360-degree view of the customer while different parts of the business are still working from different definitions of what a customer actually is. The ambition is sound and the technology may be excellent, but the answer is contested before anyone can act on it. On the other side is the organization with a mature catalog and a genuinely capable governance function that then struggles to show which governed datasets are supporting real decisions. Between them sits the relationship Atrium is arguing for: strategy tells governance where effort should be focused, and governance tells strategy what can genuinely be trusted and delivered.

It is worth being clear about something, because it is easy to misread. Atrium does not present governance as a static, documentation-led or compliance-driven exercise. It explicitly describes governance as operational and continuous, covering ownership, definitions, access controls, quality, lineage and enforcement. That matters for what follows, because what follows is not a correction of Atrium.

Atrium also makes an important connection to AI. Weak definitions, inconsistent data, unclear ownership and poor controls do not disappear when information is fed into a model. If anything, AI amplifies them, because an answer generated quickly and presented confidently is much harder for a business user to recognize as unreliable. So Atrium's argument gets us to an important point: strategy establishes what the organization wants data to achieve, while governance makes the information supporting that strategy trustworthy enough to use. Where I would extend the argument is what happens next. As trusted data moves beyond dashboards and reports into applications, copilots and AI agents, governance being strong at source is no longer enough. That does not replace the governance Atrium describes. It operationalizes it at the point of consumption.

The Gap Between Governance and the Decision

The reason organizations invest in data is ultimately quite simple. They want to make better decisions: which customers need attention, where inventory should move, whether an aircraft is ready to operate, which supplier presents a risk, where capital should be allocated. The platforms, governance programs and analytics capabilities all exist because somewhere in the organization somebody needs to decide what to do next.

This connects to a principle that has shaped the Data Tiles approach from the beginning: start with the decision and work backwards. Rather than beginning with how to govern all enterprise data, begin by asking which decisions matter, what information those decisions require and what needs to be true about that information for the business to trust it. That immediately gives governance a purpose. It tells us which definitions matter, what quality is required, how current the information needs to be, who should be able to access it and what evidence we may need later to understand how a decision was reached.

Most large organizations are not starting from nothing. They have spent years investing in warehouses, cloud platforms, integration, catalogs, governance tools, BI and increasingly AI. Yet when someone needs trusted information for a specific decision, the final step can still involve a request to another team, reconciliation across systems, another dataset or a place in an engineering backlog. Architecture diagrams rarely show this, because everything on them appears connected. But the person making the decision does not experience the architecture diagram. They experience whether they can get an answer they trust, whether they understand what it means, whether they are permitted to use it and whether it arrives in time to act. That is the final mile of enterprise data, and it is increasingly where the conversations I have with enterprise leaders end up.

That is the distinction behind Data Tiles' broader Decision-Driven philosophy. Being Data-Driven can still mean accumulating more data, dashboards and governance processes. Being Decision-Driven means working backwards from the outcomes that matter and making sure the right trusted information can reach those decisions consistently.

From Governed Data to Reusable Business Understanding

This is where the business-led data product approach we have developed at Data Tiles becomes relevant. A useful data product is not simply another dataset with a new label. It brings together the information required for a particular business purpose with the meaning, ownership, quality expectations, lineage, policies and controls needed to use it confidently. It represents an agreed understanding of something the business needs to know, packaged so it can be reused rather than reconstructed every time the same question appears somewhere else. Atrium touches on this when it discusses prioritizing the data products worth building, which is the natural bridge between strategy, governance and use.

Consider something as apparently straightforward as customer profitability. Finance may need it for planning, sales may use it when prioritizing accounts, an executive may consume it through a dashboard and an AI agent may eventually use it to recommend an action. If every one of those consumers independently reconstructs what “customer,” “profitability,” “revenue,” “period” and “cost” mean, the organization has not solved the underlying problem. It has created multiple interpretations of the same business concept and then attempted to govern each one. The alternative is to establish the business meaning, sources, ownership, policies, quality requirements and context once, and reuse that trusted understanding across approved consumers. The same applies to supply risk, workforce capacity, operational readiness, regulatory exposure and thousands of other business concepts.

McKinsey's June 2026 work on AI data readiness points the same way, arguing that scaling AI requires structured and unstructured information to become part of a governed, reusable data foundation with metadata and lineage, and recommending that leaders treat information as governed data products with defined quality thresholds and explicit rules for sensitive content. That is significant because it moves data products beyond an architectural concept and connects them to the repeatability and trust enterprise AI depends on.

Reuse creates another responsibility. If an approved definition, access policy, sensitivity classification or business rule applies when the product is created, it should not disappear because the product is subsequently consumed somewhere else. Reusability without portable governance simply moves the original risk further downstream. That is where governance needs to become active.

Governance Has to Become Active

There is an important difference between establishing governance and ensuring it remains active when data is consumed. A policy recorded in a catalog, an approved definition stored as metadata or an access rule established at source is valuable, but the real test comes later: does that governance still apply when the information reaches a business user, an application, a model or an AI agent? This is where Atrium's argument intersects with an approach we have been developing at Data Tiles: Active Governance. Governance remains connected to the trusted data product as it moves into approved consumption. Policy stays enforceable. Business context stays attached. Permissions and controls stay relevant to the consumer and the purpose. Trust is applied at runtime rather than only documented upstream.

The wider market is moving in the same direction from several angles. In its June 2026 Top Trends for Data and Analytics, Gartner identifies decision governance as an emerging requirement, because AI agents are increasingly executing strategic, tactical and operational decisions that must remain explainable, auditable and aligned to intended outcomes. Gartner's research on becoming AI-first adds a different point: those organizations need high-quality, trusted and context-rich data accessible to both humans and AI agents, which makes explicit that technical access is not the same as readiness. McKinsey's work on trust in the age of agents contributes a third, illustrating how the same agent can require different policies depending on the use case, the sensitivity of the data being accessed and the access appropriate to that situation. Policy, in other words, is becoming contextual: who or what is consuming the data, for which purpose, under which conditions.

There is an important distinction between putting governance around AI and putting governed data into AI. Enterprises need both. AI governance establishes how models and agents behave, where they can operate and who remains accountable. Active Governance addresses the information those systems are permitted to consume. Governing the model does not resolve weak definitions, uncertain lineage, inappropriate access or missing business context in the data supplied to it.

That is a subtle but consequential shift in how we think about AI-ready data. Data can be perfectly accessible and still be ambiguous, inappropriate or unsafe to use. AI-ready data needs business meaning, trust, context and governance as well as technical availability, and it matters more as the consumer changes. An analyst can notice that a number looks unusual and pause. An agent may combine it with other sources and initiate an action at machine speed. The challenge is therefore not simply giving AI more data. It is giving AI the right data, with the right business understanding, under the right policies, for the right purpose.

As governance follows trusted data further toward the decision, another question emerges: what happened after the data was prepared? Which trusted product was consumed? Which policies applied? Who or what accessed it? What recommendation followed? What decision was made?

At Data Tiles, we describe that progression as Decision Lineage. It is our term rather than one used by Atrium, Gartner or McKinsey. The objective is not another layer of documentation. It is to preserve the chain between the governed information an organization has trusted and the decision that information helped produce, which becomes particularly important when AI is part of the chain and explainability cannot stop at the model.

Governance that stops at the governed table has not yet reached the decision it was built to protect.

The APJ Perspective

This challenge takes on additional dimensions across Asia Pacific and Japan, because there is no single APJ data environment. Organizations operate across markets with different regulatory requirements, sovereignty expectations, technology maturity, languages, customer behaviors and operating structures. A global business definition may require local interpretation, a policy set centrally may need to be applied differently in Australia, Singapore, Japan, Indonesia or India, and data may need to remain within a jurisdiction while still contributing to a regional or global decision. Most enterprises are navigating that complexity across technology estates built over decades rather than clean architectures designed yesterday.

That reality makes wholesale replacement impractical. The answer cannot be to wait until every platform has been modernized, every definition standardized and every governance program completed before the business is allowed to move, and effective governance should not require centralizing everything first. Different jurisdictions can retain the infrastructure, controls and policies they require while the enterprise creates a consistent mechanism for trusted data use. Policies can reflect local requirements. Access can reflect the individual consumer and purpose. Business context can remain attached even when the underlying data stays where it already lives. The objective is not to make every market identical. It is to make trusted understanding portable while allowing the appropriate governance to travel with it.

There is also very little appetite among the enterprise leaders I speak with for another multi-year promise that begins by replacing everything the organization already owns. The questions are far more practical. How do we get more value from the platforms we already have? How do we make governance useful to the business rather than another process to navigate around? How do we prepare enterprise data for AI without building a parallel data estate? And how do we do that across multiple markets without pretending every market operates the same way?

Those are healthy questions, because they move the conversation away from technology acquisition and toward business use. The organizations making progress in the region do not begin by trying to perfect their entire data estate. They identify important decisions, understand the information those decisions require, apply the appropriate governance and business context, and make that trusted understanding reusable. Then they move to the next decision, and the capability compounds.

The Market Signal

Atrium is right to distinguish data strategy from data governance, and equally right that organizations need both. Strategy determines what the enterprise wants its data to accomplish. Governance establishes the trust required to accomplish it. The next stage of the conversation is what happens when that trusted data reaches consumption and the decisions it was meant to support.

Different industry voices are approaching that problem from different directions. Gartner is arriving at it through decision governance and through the trusted, context-rich data humans and agents both require. McKinsey is arriving at it through governed, reusable data foundations and contextual policy around agents. Collectively, governance is moving closer to consumption and decision-making. Our contribution to that discussion is Active Governance.

This is the problem we have been focused on at Data Tiles. Most enterprises do not need another replacement data platform. They have already invested heavily in platforms, governance, integration, analytics and AI. The missing piece is making those investments useful at the point where the business actually needs to decide. Latttice enables business teams to create trusted, governed, fit-for-purpose and reusable data products over the technology the enterprise already has, with governance and business context remaining connected as those products move into consumption.

Lenz extends that trusted foundation into AI, so agents work from governed, business-led data products rather than independently reconstructing an understanding of the organization for every use case.

The enterprise does not invest in data because it wants more data. It invests because better information should lead to better decisions and, ultimately, better outcomes. AI did not create the final-mile problem between enterprise data and enterprise decisions. Business teams have lived with it for years. AI makes leaving it unresolved much harder to ignore.

Strategy tells us what matters. Governance establishes the rules and the conditions for trust. Active Governance keeps those rules connected as data moves into use. Business-led data products make trusted business understanding reusable. People and AI consume that trusted understanding. And the decision is where the value is ultimately realized.

Strategy without execution remains a plan. Governance without use remains a control. Their value appears when trusted data reaches a decision.

The destination of governance is not the governed table. It is the trusted decision.