Research Report
AI-ready data
New rules of data for the advanced AI era
10-minute read
May 26, 2026
In brief
- AI momentum is outpacing data-readiness, as siloed, low-quality data & lack of business context & tacit knowledge inhibit safe, scaled AI adoption.
- Only 7% surveyed—data reinventors—have progressed far enough in building AI-ready data capabilities required for scaled AI adoption of advanced AI.
- Data reinventors realize significant value, with profit margin uplifts up to 1.6x over industry peers.
The data-readiness gap
Businesses are moving from experimenting with advanced AI to enterprise-wide adoption. Yet, our research, spanning proprietary modeling, expert interviews and a survey of executives at 2,000 companies in 15 countries and 9 industries, shows a clear need to reinvent data foundations to support that ambition.
64%
of survey respondents say their businesses have moved beyond pilots into production across multiple functions or initiated coordinated efforts enterprise-wide for advanced AI.
7%
But our analysis shows that only 7%—data reinventors—have reached the level of data-readiness required to scale advanced AI, which includes generative, agentic and physical AI.
What is AI-ready data?
Data is already widely used to support decisions, but as advanced AI gains traction, the next step is to let AI systems initiate actions. This requires a fundamental shift in how organizations manage data.
Foundational properties such as completeness, accuracy, consistency, accessibility and governance remain essential but are no longer sufficient. They were built for structured tabular data meant for use by human experts like data scientists and software engineers for reporting and traditional machine learning. Much of this enterprise data is static and siloed. Advanced AI is now exposing limitations such as poor data quality and fragmentation faster than any previous technology.
To be AI-ready today, organizations still need trusted, governed data that meets the requirements of traditional AI. But advanced AI demands richer context. Data must also serve as a referential corpus drawing from both structured sources and unstructured ones like documents, diagrams and code that capture processes, experience and know-how. It must include the semantic meaning of common terms, entities and their relationships as it applies to a process set as well as comprehensive examples that illustrate use cases. And, increasingly, it must be refreshed in real time.
- Most organizations are not there today. 72% do not have trusted data of the right quality, overlaid with standardized governance practices to support advanced AI.
- More than 80% delay, limit or alter AI and gen AI initiatives at least occasionally because of data-related risks.
How data reinventors turn readiness into value
Our research shows that data reinventors, the leaders in AI-ready data capabilities, realize more value across both financial and non-financial outcomes such as productivity, decision quality, customer experience, risk reduction and margin improvement.
- They have an estimated 4.5 percentage point advantage over industry peers on earnings before interest and taxes (EBIT) margins, the equivalent of a margin uplift of up to 1.6x in the last three years.
These organizations are much more intentional in focusing resources on strategic bets—a growing set of high-impact initiatives in core operational areas—rather than spreading effort across lower-value, incremental work.
- They direct data efforts toward core domains in their respective industry value chains such as research and discovery in life sciences, network and service operations in communications or core banking operations.
- They are nearly 2x more likely than their peers to focus resources on these areas.
A practical and continuous path to AI-ready data
Data reinventors build AI-ready data through a continuous, iterative loop rather than a one-time transformation. They focus on strategic initiatives where better data can create clear business value, then use a minimum viable product mindset to test, scale and improve as they go.
Value comes from identifying the right information, giving it shared meaning for both people and AI systems and translating it into measurable business outcomes.
A global hotel chain illustrates this. For years, it managed bookings through an automated reservation system that treated rooms as broad categories, capturing guest preferences such as floor, view or balcony as free-text notes. This made specific requests hard to honor or price. By restructuring these attributes as bookable data fields, the chain improved experience and unlocked incremental revenue.
The path to value unfolds across three interconnected stages.
01
Capture trusted, reusable and operationally ready data.
To overcome siloed and unreliable data, organizations must first re-architect how they capture it and connect it across the enterprise.
Key actions
1. Establish a federated, hub-and-spoke data model in which domains own the value, context and quality of their data, while a central hub sets common standards, talent models and engineering patterns.
While centralizing data in one cloud environment might be most efficient, this is not a realistic ambition, given budgets, changing priorities and data sovereignty. It is more pragmatic to aim for a single logical view of data that remains federated. Organizations can achieve this by applying common enterprise standards with interfaces for data security, quality and access, while enabling distributed use across domains, using AI and agents to modernize and activate only the data that matters most to priority use cases.
1.3x
data reinventors are more likely to operate with federated data and active metadata than the rest.
65%
adoption among data reinventors and 50% among the rest.
6%
of all organizations have a unified logical data view across systems and ecosystems today: 30% of data reinventors and 4% of their peers.
2. Put appropriate data governance in place so that data is secure, of high quality, compliant and fit for use by advanced AI. This means embedding policies, provenance, validation and compliance checks into data workflows, so trust is built in right from the start.
95%
of data reinventors have a cohesive data management and governance system, compared to 49% of others.
1.8x
more likely to achieve scaled and automated data management and governance in the next 18 months: 79% versus 43% of their peers.
3. Productize data by making structured, unstructured, real-time and synthetic data explicit, reusable and interpretable by both human and machine users.
3x
data reinventors are more likely to intentionally engineer reusable data products, with 65% doing so compared with 22% of others.
02
Contextualize the data.
Giving data shared meaning and context across advanced AI systems is vital. Existing systems of record still matter, but systems that own the why behind work data will matter even more. This domain knowledge resides in workflows and lived experience and has a critical role to play in agentic systems’ moments of truth like exceptions, approvals, cross-system synthesis and precedent.
Key actions
1. Build a semantic layer, so humans and AI can reason and act using the same business meaning. Domain ontologies, deployed as knowledge graphs, connect data with context in a repeatable and explainable way, helping ensure the right data is used for the right purpose.
11x
data reinventors are more likely than the rest to convert tacit, expert-driven knowledge into maintained, reusable and continuously improving enterprise intelligence.
2. Establish context. Capture the why, what, how and now what beyond systems of record. Context graphs, for instance, recall and trace why the business acted within agentic systems. Systems of record capture the what of the business: transactions, customers, products, assets and processes. Advanced AI also needs the how, which sits in unstructured knowledge such as documents, policies, workflows, communications and expert judgment, as well as the now what from real-time signals to respond to change and recommend or initiate the next best action.
74%
of data reinventors embed decision intelligence across multiple core business decisions, enabling consistent, governed actions, compared with just 28% of industry peers.
2x
more likely to deploy context graphs at scale compared to their peers; almost 50% of data reinventors have done so versus 23% of the others.
03
Realize value from the AI‑ready data journey.
As data quality improves, it strengthens semantic reasoning. Stronger semantics in turn enable sharper questions and richer answers, creating greater business value. That value drives teams and ecosystem partners to further refine and clarify the data, reinforcing the cycle.
Key actions
1. Establish a two-speed value realization model for AI-ready data initiatives. Prove priority AI use cases quickly, in weeks rather than months, using minimum viable data products to test the value case. Once value is demonstrated, use the results to guide deeper investment in data quality, governance, context and reuse so the solution can scale reliably.
- Treat outcomes not as endpoints but as signals for where to improve data quality, deepen context, refine governance and focus on the next strategic bets.
- Establish guidelines that connect quick wins to long-term investment. These could include incentives for enterprise-wide use of data products rather than local optimization, metrics tied to reuse, value attribution for contributing domains and central funding for horizontal data products. This helps avoid a common trap: achieving strong pilot performance but struggling to scale across the enterprise.
2. Infuse AI into the data lifecycle to accelerate and improve data quality. Use it as the operating layer for managing, governing and activating data end-to-end. Gen AI and agents can automate repetitive tasks such as metadata tagging, lineage mapping, quality monitoring, policy checks and documentation, while shifting human roles toward judgment, validation, exception handling and decision oversight.
95%
of them use AI for data mapping and lineage, 95% for metadata enrichment, 90% for data quality monitoring and 84-90% for conversational reporting and insights generation.
3. Activate the partner ecosystem to extend the organization’s capabilities, bring in external talent, data and intelligence and accelerate innovation, interoperability and scale.
81%
of data reinventors plan to increase investments in AI‑specific enablers, 79% of hyperscalers and 61% of specialist data platforms within the next 18 months, strengthening the foundation for scalable ecosystem participation.
3.3x
data reinventors are more likely to actively participate in broader ecosystems or data marketplaces with multiple partners, shared standards and reuse across use cases. 82% of data reinventors do so versus 25% of the rest.
Make data a compounding advantage
AI-ready data is fast becoming a decisive enterprise advantage. Competitive edge will depend less on having more of the same data. Real breakthrough value comes from metadata and overlooked signals, like tacit knowledge, making data trusted, understandable to humans and systems, easily accessible, and exposing previously unsolvable, high‑value business problems.
Our research shows that data reinventors are already turning this shift into measurable value by following a continuous path: capturing trusted and reusable data, contextualizing it with shared business meaning and realizing value in ways that fund and sharpen the next cycle.
From this path, the following imperatives stand out for all organizations now:
Capture
Federate
Productize data
Extend data-readiness
Realize value
Data reinventors are already capturing outsized advantages quickly, solving previously unsolvable business problems, and making it harder for their industry peers to catch up later.
Coming soon: Our new report on data reinventors, their success stories, the capabilities that set them apart and the actions others can take to close the gap.
More on this topic:
Tacit Knowledge Is Your Next Competitive Moat | California Management Review
Data essentials in the age of generative AI
