Strategic Clarity Has to Reach the Decision.

A response to BCG's Applied AI Index 2026 · By Cameron Price, Founder & CEO, Data Tiles

There has been no shortage of debate about whether enterprise AI is actually creating value. Boston Consulting Group’s newly released Applied AI Index 2026 makes an important contribution to that conversation because it looks beyond individual pilots, demonstrations and anecdotes and asks what is happening across more than 1,300 organizations. The findings are encouraging. AI is beginning to create measurable value at scale. But BCG also shows very clearly that simply having access to AI is not what separates the organizations creating meaningful value from everyone else. The difference is increasingly about clarity: clarity about where AI should create value, which business processes are worth changing, what capabilities are required and how success will ultimately be measured.

What BCG's Applied AI Index Tells Us

BCG’s research, based on a survey of 1,330 CxOs and senior leaders across more than 20 sectors, finds that almost half of the organizations studied are now generating value from AI. BCG classifies 7.5% as “future-built,” 41% as “scaling,” 47% as “emerging” and 4.5% as “stagnating.” The performance gap between those groups is substantial. The most advanced organizations achieve 2.3 times the total shareholder return, 2.4 times the revenue growth and 2.8 times the EBITDA growth of laggards. Perhaps just as important, another 41% of organizations are now in BCG’s scaling category. AI value is no longer confined to a handful of exceptional organizations. It is beginning to broaden.

Strategic Clarity + Applied AI

The question, then, is what separates the organizations moving forward from those still struggling to turn AI investment into business outcomes. BCG summarizes its answer simply: Strategic clarity + Applied AI = Transformative impact. Organizations combining strong strategic clarity with mature applied AI capabilities report AI-generated value equivalent to 4.5% of revenue, compared with 0.9% for organizations weak on both dimensions. That is the five-times difference highlighted in the Applied AI Index. But the other two numbers are equally interesting. Mature AI capabilities without sufficient strategic clarity deliver 2.0%, while strategic clarity without the capabilities required to execute delivers 1.8%. Neither technology nor strategy is enough on its own.

BCG describes strategic clarity in practical business terms. Organizations need to understand where competitive advantage is shifting, identify a relatively small number of high-value opportunities, connect initiatives across the enterprise and measure whether those investments are producing the intended outcomes. Among BCG’s future-built organizations, 95% use clear KPIs or directly track P&L value from AI. Organizations directly tracking AI through the P&L generate three times the AI value of those that do not formally measure it, 3.6% compared with 1.2%. That is an important distinction because the objective was never really AI adoption. The objective is a better business outcome.

AI Spending Is Moving Into the Business

There is another finding in the Index that I think tells us a great deal about where this market is heading. BCG reports that AI spending has risen to approximately 3.3% of organizational revenue, more than tripling from around 1% in early 2025, but 80% of AI spending now sits outside the enterprise IT budget. AI is becoming less of a technology initiative that the business consumes and more of a business capability that technology enables. BCG’s future-built organizations reinforce that point. They are 3.5 times more likely than laggards to fund AI through a single multiyear program, and they concentrate investment on fewer, higher-value workflows rather than simply distributing AI across the organization because the technology is available.

From Assistance to Agency

That brings us to what I think is one of the most consequential parts of the research: the move from AI assistance to AI agency. BCG estimates that agents represented 17% of total AI value in its 2025 sample, rising to 22% in 2026, and projects that this could reach 39% by 2030. Among future-built companies, 44% already report realizing value from agentic AI, compared with only 2% of laggards. But increasing autonomy changes the problem. By 2030, 42% of the companies surveyed expect AI agents to operate autonomously, making decisions without human approval. Today, only 5% have the full set of critical controls in place.

That gap should get our attention. We are talking about giving machines considerably more authority over business processes while most organizations are still working out the controls required to govern that authority. BCG’s evidence also suggests that governance should not simply be thought of as a brake on innovation. Organizations with all six of the AI controls examined by BCG generate three times as much agentic AI value as organizations with only one control in place. In other words, good governance and greater autonomy do not have to be opposing forces. Done properly, governance is part of what makes greater autonomy possible.

The Data Underneath the Agent Matters

There is another part of BCG’s framework that particularly interests me. One of its three Applied AI capabilities is to build an agentic platform powered by data. The emerging technology gap, in BCG’s view, is not simply access to better models. Organizations need enterprise intelligence and platform layers that allow agents to understand context, take governed action and operate reliably at scale. Among future-built organizations, 95% are undergoing a data transformation, while more than two-thirds are moving toward an enterprise-wide AI platform with common architecture and a control plane rather than relying on a single-vendor stack.

And this is where I think the conversation becomes particularly interesting.

An AI model can be extraordinarily capable and still be poorly equipped to make a particular business decision. It does not automatically know what a number means inside your organization. It does not necessarily know which definition of “customer” applies, whether a particular data source is appropriate for the decision being made, how one piece of information relates to another, which policy applies, who is permitted to see something, how current that information needs to be or what evidence should accompany a recommendation. A person who has worked inside that business for years often carries much of that context implicitly. An AI agent does not.

Context therefore becomes part of the infrastructure of enterprise AI.

For me, this is where BCG's findings begin to point to a wider market signal. The AI value conversation is moving beyond access to intelligence and toward the quality of the decisions that intelligence enables. And BCG is not the only organization seeing the underlying requirements change.

Gartner has increasingly described context as critical infrastructure for AI. In its 2026 work on becoming an AI-first organization, Gartner argues that high-quality, trusted and context-rich data needs to be accessible to both humans and AI agents, with semantics, metadata and knowledge providing the context agents need to reason and deliver trusted intelligence. Gartner has also reported that organizations with the highest maturity in AI-ready data and analytics capabilities are achieving up to 65% greater business outcomes.^[2] The message is important: success is not simply about better models. It is increasingly about giving AI governed, contextual access to the right data.

Recent Gartner research on data products takes that idea further. Its July 2026 playbook for agentic AI-ready data products argues that data product practices need to evolve to support both human and machine consumption, including changes to governance, active metadata and machine-verifiable data contracts.^[3] In August, Gartner described the next evolution even more directly: AI-ready data needs to expand to agent-ready data as organizations begin deploying systems capable of reasoning and acting with greater autonomy.^[4]

I do not see these as separate trends. Taken together with BCG's findings, they point toward a fairly fundamental change in what enterprise data needs to do. It is no longer enough for data to be available. It increasingly needs to arrive with enough meaning, context, evidence and governance for a person or machine to use it appropriately in a particular business situation.

Strategic clarity has to reach the decision.

Taken together, I think these signals change the question organizations need to ask.

During the first phase of generative AI, much of the market understandably concentrated on the model. Which model? Which assistant? Which platform? Which use case? Those questions still matter, but the questions that determine value are becoming different. Where do we actually want AI to create value? Which workflows matter? Which decisions within those workflows need to improve? What context does AI need to participate in those decisions? What evidence should it be able to rely upon? What controls need to remain attached when AI begins to act?

That becomes even more important as AI moves from answering questions to taking action. Gartner's 2026 Data and Analytics trends identify decision governance as an emerging requirement as AI agents execute more strategic, tactical and operational decisions. Gartner argues that automated decisions increasingly need to be explainable, auditable and aligned with intended outcomes.^[5] BCG arrives at the issue from another direction, finding that 42% of organizations expect agents to operate autonomously by 2030 while only 5% currently have the full set of critical controls in place.^[1]

That is the signal I think matters.

The model is becoming more capable. The agent is becoming more autonomous. The business is becoming more involved. And the data underneath all of it now has to carry far more than a value in a field. It needs to carry the context that makes that value useful, and the governance that determines how it can be used.

That is a substantially more mature conversation, and it leads directly to a principle we have increasingly adopted at Data Tiles: start with the decision, then work backwards.

Start with the decision. Then work backwards.

If an organization wants AI to improve an outcome, begin by understanding the decisions that create that outcome. What decision needs to be made? What context does the person or agent need? What evidence can be trusted? What policies apply? Who should have access? What needs to be explainable afterwards? Then we can ask how the technology delivers it. This is the thinking behind the Decision-Driven Enterprise.

1. Business Priority
2. Workflow
3. Decision
4. Context
5. Evidence
6. Governance
7. Action
8. Outcome

BCG talks about strategic clarity at an organizational level. I think there is an important extension to that idea at an operational level. Strategic clarity has to survive the journey from the boardroom to the individual decision. An organization can be absolutely clear that reducing claims leakage, improving demand forecasting, increasing aircraft availability or detecting fraud is strategically important. But the person or AI agent operating inside that process still needs the right context at the moment the decision is made.

There is a chain between strategy and value: business priority → workflow → decision → context → evidence → governance → action → outcome. Break that chain anywhere and strategic clarity can be lost before it becomes business value.

That is why I do not think the next stage of enterprise AI will be solved simply by putting more organizational data within reach of increasingly powerful models. The objective should not be to give AI access to everything. It should be to give AI the right governed context for the decision it is being asked to support.

From Market Signal to Practical Capability

This is also where the conversation becomes very practical for us at Data Tiles, because much of what this research is now describing is the problem we have been building Latttice to solve.

We did not set out to build another AI model or another place to store enterprise data. We built Latttice around a much simpler question: how do we get trusted data, with the context and governance required to use it, closer to the business decision?

That starts by reversing the traditional journey. Instead of beginning with the data and asking what we might do with it, start with the decision and work backwards. What decision needs to be made? What context does it require? What evidence can be trusted? What governance needs to apply? Then bring the business understanding and the enterprise data together around that need.

Business knows the decision. Data knows the evidence.

Latttice provides the workbench where those two things can come together.

A claims specialist understands the decisions involved in a claim. A maintenance manager understands what determines whether an aircraft should return to service. A supply-chain planner understands the context behind a late shipment. A clinician understands the information surrounding a patient decision. That domain expertise is valuable context. The opportunity is to combine that understanding with governed enterprise evidence without asking every domain expert to understand the technical architecture underneath it.

With Latttice, business and data teams can create governed data products around the context required for a decision without asking business users to become data engineers. Latttice connects to the enterprise data organizations already have and allows the business understanding around that data to become part of the product itself: meaning, context, policy, access, lineage and the evidence required to use it appropriately.

The engineering still matters. The architecture still matters. Security and governance absolutely still matter. The difference is that the technical complexity does not have to become the business user's problem.

The technical complexity stays under the hood. Complexity is abstracted. Governance is not.

Once that governed context exists, it should not have to be rebuilt every time the consumer changes. The same governed data product can support a person making a decision, a BI dashboard, an application, a workflow, an AI assistant or an autonomous agent.

Create once. Govern once. Use everywhere.

What makes this particularly interesting to me is that organizations do not necessarily need to wait for another wholesale data transformation before they begin working this way. Latttice is designed to work with the data infrastructure and governance investments organizations already have. The underlying environment can continue to evolve while the business begins creating governed, decision-centric data products now.

That changes the sequence. Instead of waiting for every data problem to be solved before AI can be trusted, organizations can start with a valuable decision, assemble the governed context required for it, prove the outcome and reuse that same governed product across human and machine consumers.

We don't replace your data platform. We complete it.

Latttice provides the governed layer that brings trusted data and business context closer to the point of decision, while the underlying data platforms, governance investments and infrastructure continue to do what they were built to do.

Active Governance When the Consumer Can Act

That becomes particularly important when we return to BCG’s finding that 42% of companies expect agents to have autonomous decision-making authority by 2030 while only 5% currently have the full set of critical controls in place.^[1] When the consumer of the data can reason and act, governance needs to travel further than the catalog. It needs to apply when the governed product is created and when it is consumed. What is this agent allowed to see? Which rows and columns are appropriate? Which policy applies to this particular decision? Is the information sufficiently current? Can we identify where it came from? Can we explain the evidence behind the action?

This is why I think governance has to move closer to the decision. Cataloging, classification, ownership and lineage remain essential, but they are no longer the end of the governance journey. When the consumer can reason, recommend and act, governance also needs to determine what information can be used, by whom or by what agent, under which conditions and with what evidence. Gartner's emerging focus on decision governance reinforces that direction.^[5] The question is increasingly not simply whether the data is governed somewhere upstream, but whether governance remains active when that data becomes part of a decision.

That is what we mean at Data Tiles when we talk about Active Governance. Governance cannot simply describe the data somewhere upstream. It increasingly needs to operate at the point of consumption, particularly when the consumer is no longer necessarily a person.

Governance cannot simply describe the data somewhere upstream. It increasingly needs to operate at the point of consumption.

Explore Active Governance