The CEO’s AI Dashboard, Part 3: Data Readiness
Welcome to a new edition of The Board: Distillation Aftershots (*).
This online post is a copy of the newsletter, written by Constellation's Chief Distiller and Board Advisor Esteban Kolsky, that shares curious and interesting insights and data points distilled from enterprise technology to identify what’s notable. To subscribe, please click here.
In this issue, I want to move from business value to the condition that usually determines whether that value appears at all: data readiness. Enterprises still talk about data readiness as though it were a cleanup project. It is closer to an operating discipline. Good data helps. Clean data helps. The right data, in the right place, with the right meaning attached, matters more.
First, my take.
Before you can use AI well, you need data the enterprise can trust well enough to act on. That means data flows, labels, semantics, and controls that make the data usable in context. Case studies from early AI deployments show that projects stall because data is fragmented, poorly governed, semantically inconsistent, or too slow to move where it needs to go. Bain states this directly: most AI pilots do not scale to full production, often because of poor data quality, unclear ownership, and inconsistent governance.
Organizations are slowing AI adoption behind large data-cleanup efforts. Those projects are often too slow for the moment. If the goal is to learn quickly, improve early results, and redesign the business around new workflows and processes, then “perfect data later” is often a weaker strategy than “relevant data now.” Readiness is the ability to get the right data to the model, the workflow, and the user at the right time, with enough context attached to make the output useful.
I would treat data as a strategic asset class, not as a byproduct of applications. IBM’s 2025 CDO study is useful here: 78% of CDOs say leveraging proprietary data is a top strategic business objective for differentiation, 81% are prioritizing investments that accelerate AI capabilities, and nearly half say advanced data skills are now a top challenge, up from 32% in 2023.
That is a clear signal that enterprises already understand the direction of travel.
The enterprise’s most valuable data assets are the combinations competitors cannot easily replicate: transaction history, telemetry, usage patterns, outcomes, classifications, and domain semantics accumulated over time. That is where privileged data becomes a moat. Data ownership should increasingly be defined by workflow, not just by system or function; AI exposes value across end-to-end work, and the data has to be organized the same way.
Data readiness is not only about quality or availability, but also about understanding what the data means, what it is called across the enterprise, where it originated, what context it requires, and how it can be used safely. Semantics matter more than most enterprises admit. PwC’s data-governance guidance is explicit on this point: establish a single source of truth for high-value data assets, standardize data definitions and access protocols, and invest in interoperability, lineage tracking, and metadata management to make AI explainable and usable.
Ontologies, taxonomies, data dictionaries, lineage, provenance, and classification engines sound like old data-management terms. They are becoming current again because AI systems quickly expose the cost of semantic confusion.
There is also a confidentiality issue emerging here that enterprises should take more seriously. Privacy remains essential, but privacy is no longer the full boundary. Data confidentiality is becoming an increasingly operational question: which data can move, which data can be inferred, which data can be exposed to agents, and which data must remain constrained even within approved systems. If agents are going to operate across workflows and systems, enterprises need confidentiality engines and access strategies that protect data integrity not only from the agent layer but also from external misuse.
Finally, if the enterprise cannot move its enriched data, labels, semantic layers, and workflow intelligence across vendors and architectures, then part of its moat is sitting inside someone else’s platform. That is one more reason private platforms matter. They are not only an infrastructure choice. They are a control point for keeping proprietary data, meaning, and iteration logic under enterprise governance rather than allowing them to drift into vendor dependence.
The operating side of this is already visible. PwC’s 2026 Digital Trends in Operations Survey of 767 U.S. operations and supply chain leaders found that 85% believe they are ahead of competitors in digital transformation, yet 89% say their technology investments have not fully delivered expected results. That gap is a useful warning. Confidence in digital progress is not the same as readiness to use data well across AI-driven workflows. Most enterprises have enough data to explain yesterday. Far fewer have data ready to support a decision, trigger a workflow, or guide an agent safely today.
Companies that understand data readiness are not using AI to optimize the business exactly as it works today. They are using AI to redesign parts of the business and ensuring the data infrastructure can support those redesigns. That is where the moat compounds: when each workflow produces more useful data, and the enterprise can capture, govern, and reuse it faster than competitors can match.
The CEO question is not whether the enterprise has clean data.
The CEO question is whether the enterprise has the data flows, semantics, controls, and confidentiality to make AI useful where the business is going next.
If the answer is no, more model access will not fix it.
Recommended CEO actions:
- Identify which proprietary data flows could compound advantage if captured, labeled, and reused across redesigned workflows.
- Fund the semantic layer, not just the data platform: ontologies, taxonomies, data dictionaries, lineage, classification, confidentiality rules, and access controls.
- Treat private platforms as the operating answer to data portability, control, and iteration risk. If the enterprise cannot migrate its enriched data and semantic assets when vendors or architectures change, the moat is weaker than it appears.
Here are some reading resources:
- IBM’s 2025 CDO study is useful because it finds that AI ambitions are outpacing enterprise data readiness and that proprietary data is increasingly viewed as a differentiator.
- IBM’s 2026 Tech Leader Study is useful because it frames AI scale as a foundation problem, with structural readiness, governance by design, and infrastructure adaptability at the center.
- PwC’s responsible AI and data governance piece is useful because it emphasizes lineage tracking, metadata management, interoperability, and standardized definitions for AI-ready data.
- PwC’s 2026 digital trends in operations survey is useful because it connects AI performance to modernized data foundations and redesigned operating models rather than model access alone.
- PwC’s AI agent survey is useful because it shows that one of the biggest barriers to value is connecting agents across applications and workflows, which fits the point that data flow matters more than cleanup theater.
- Bain’s “Why AI Stumbles Without a Solid Data Strategy” is useful because it argues directly that most AI pilots fail to scale when data quality, ownership, and governance remain weak.
- Bain’s 2026 piece on accelerating AI transformation is useful because it argues that high-value AI opportunities typically lie where rich proprietary data and critical workflows intersect.
- Bain’s 2025 SaaS disruption piece is useful because it makes the proprietary-data argument plainly: the moat sits in usage patterns, domain-specific content, and transaction history, not in common models.
What’s your take? We are fostering a community of executives who want to discuss these issues in depth. This newsletter is but a part of it. We welcome your feedback and look forward to engaging in these conversations.
If you are interested in exploring the full report, discussing the Board’s offering further, or have any additional questions, please contact me at [email protected], and I will be happy to connect with you.
(*) A normal distillation process produces byproducts: primary, simple ones called foreshots, and secondary, more complex and nuanced ones called aftershots. This newsletter highlights remnants from the distillation process, the “cutting room floor” elements, and shares insights to complement the monthly report.
Esteban Kolsky spent the first part of his professional career working on the customer service front line and in management roles, culminating in early implementations of first-generation CRM tools. In 2000, he became a research analyst at Gartner, focused on CRM, customer service, customer experience management, enterprise feedback management, and voice of the customer, among topics that later became the basis for the current field of CX. During his tenure at Gartner, Kolsky assisted vendors and Global 2000 organizations with their customer strategies. He left Gartner to focus his career on strategic advisory services and influenced all the major software vendors, as well as many smaller vendors, in the CX industry and advised executive leadership teams across industries and countries…...
Read morePublished
Author
