Executive Summary
Enterprise AI has reached near-universal adoption and near-universal disappointment at the same time. McKinsey’s latest global survey shows 88% of organizations now use AI in at least one business function, yet only about 6% capture more than 5% of EBIT from it. The instinct is to blame the model. The evidence points elsewhere: the companies that win have redesigned workflows and, underneath that, built data their systems can actually trust. In 2026, AI agents move into production faster than governance and data quality can keep up, and only 15% of companies report data environments genuinely ready to run them. This edition argues one thesis: the intelligent company is not the one with the best model. It is the one with data it can trust, structured for machines and for markets. That same data now drives valuation. The Stobox view is that intelligence, investment-readiness, and capital access are one continuous build, not three.
Key Takeaways
- Adoption is mainstream but value is rare: 88% of organizations use AI in at least one function, while only about 6% attribute more than 5% of EBIT to it, per McKinsey’s State of AI 2025.
- The differentiator is not the model. High performers were nearly three times more likely to have fundamentally redesigned workflows, and only about 21% of adopters have redesigned any workflow at all.
- Data readiness is the choke point: only 15% of companies report data environments fully prepared to run AI agents in production, even as roughly 41% already deploy them.
- Governance is the new production risk: Gartner forecasts more than 40% of agentic AI projects will be cancelled by 2027, driven by unclear ROI, escalating cost, and weak controls.
- The same structured, verified, governed data that makes AI work is now what investors and acquirers price. Private equity firms are already discounting companies carrying “AI debt.”
The Adoption Is Real. The Advantage Is Not.
The direct answer: nearly every company now uses AI, but almost none turn that use into measurable profit, and the reason is structural, not technological.
Start with the numbers, because the numbers are unusually clear. AI is now mainstream. McKinsey reports that 88% of survey respondents say their organizations regularly use AI in at least one business function, and 72% report using gen AI, up from 33% in 2024. That is the adoption story, and it is finished. The interesting story is what happens next.
Almost nothing, for most companies. In McKinsey’s 2025 State of AI survey of nearly 2,000 organizations across 105 countries, 88% reported using AI in at least one business function. Only 6% qualified as high performers, companies achieving more than 5% EBIT impact from their AI use. Put differently: nearly nine out of ten companies have the tools. Fewer than one in ten have the returns.
The reflex explanation is that the models are not good enough, or that the right use case has not been found. The data rejects that. The usual diagnosis, that what is missing is a better model, more training or the right use case, does not hold, because high performers and laggards use the same models. Everyone has access to the same frontier systems. The gap sits in what those systems are pointed at and what feeds them.
This is the defining tension of the year, and it changes the executive question entirely. The benchmark question has changed. “Are we using AI?” stopped being useful in 2024. “Have we redesigned a workflow around AI and can we prove it?” is the 2026 question.
What the 6% Actually Do Differently
The direct answer: the high performers rebuilt the work and the data underneath it, rather than bolting AI onto existing processes.
Two practices separate the winners, and neither is about buying better technology. The first is workflow redesign. High performers were 3.6 times more likely to pursue enterprise-level organizational change, and 55% of them fundamentally redesigned workflows when deploying AI, compared to roughly 20% of other firms. And redesign is not common: fundamentally redesigning workflows has the strongest link to EBIT, yet only 21 percent of adopters had fundamentally redesigned any workflow.
The second practice is investment in the data layer that AI runs on. AI high performers are 3x more likely to have strong senior leadership engagement, have redesigned workflows end-to-end, set outcome-based objectives tied to business KPIs, and invest in agent-ready infrastructure. BCG frames the split cleanly: seventy percent of AI’s transformative value depends on meaningful transformation of people, organization, and processes, not on algorithms or infrastructure. The technology accounts for only 30%.
There is a useful, honest read of the underlying mechanic here. Automating a single task rarely helps for long. It is because automating tasks just moves the bottleneck to the next human-based bottleneck. The 6% avoid this by rebuilding a whole process and setting AI as the default inside it, then measuring the result against EBIT rather than against license counts. That discipline, not model choice, is the moat.
The Real Choke Point: Data You Can Trust
The direct answer: AI agents are entering production faster than corporate data can support them, and most data environments were never engineered for autonomous systems.
2026 is the year agents left the demo. AI agents have left the lab and entered production. Gartner projects that 40% of enterprise applications will have embedded agents by the end of the year, up from less than 5% in 2025. But deployment has outrun readiness. Only 15 percent of companies are fully prepared to deploy agent-based AI in production, even though nearly 60 percent report investing tens or hundreds of millions in this area.
The people building the plumbing say it plainly. Most companies fail with AI not because of the models, but because their data isn’t ready. Companies are deploying agent-based AI on top of fragile pipelines, a lack of traceability, and systems that were never designed for autonomy. Independent surveys agree on where the constraint sits. Data quality, availability and access lead the barrier list, followed by integration, skills and governance. The constraint is enterprise plumbing, more than model horsepower.
Why does bad data hurt agents more than it hurt earlier analytics? Because agents act. Traditional architectures optimized for batch processing and human-guided analytics cannot reliably support persistent, concurrent, real-time decision-making by autonomous systems. An error in a dashboard is a wrong number. An error in an agent is a wrong action, executed at machine speed, often without a human in the loop.
This is why so many projects will die. Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value, or inadequate risk controls. The projects that fail are not failing because the technology does not work. They are failing because the organizations deploying them were not ready. Governance is now the exposed edge: Arctera’s State of AI Governance 2026 report finds that 78% of organizations using AI expect communications risk to increase, while fewer than one in five can prove that their governance controls are working.
Definition: An AI-ready company is an organization whose business data is structured, verified, current, traceable, and governed to the standard required for autonomous systems to act on it safely and for external parties to trust it. It is a property of the data and the operating model, not of the model in use.
The vendors moving fastest have internalized this. SAP’s answer at Sapphire 2026 was consolidation, not a smarter chatbot: agents automatically inherit business context, data lineage, and process rules without manual configuration. The lesson generalizes. Fix the context layer first. The agent is the easy part.
A Framework: The 5 Stages of Becoming an Intelligent, Investment-Ready Company
The direct answer: readiness is a sequence, and skipping the early stages is exactly why pilots stall.
Most organizations attack AI in the wrong order, buying tools before they have data those tools can trust. The following framework maps the Stobox three-stage narrative (build intelligence, become capital-market ready, connect to digital finance) onto a practical progression.
| Stage | What it means | The failure if skipped |
|---|---|---|
| 1. Intelligence | Consolidate, verify, and structure business data into a single trusted context layer | Agents act on fragile, siloed data; pilots stall |
| 2. Workflow redesign | Rebuild a full process around AI as the default, measured against EBIT | “Many local wins, little systemic reinforcement” |
| 3. Governance and trust | Data lineage, human-in-the-loop validation, auditable controls | Cancelled projects and compliance exposure |
| 4. Investment readiness | Turn trusted internal data into investor-grade, verifiable reporting | Diligence discounts and slow, painful capital raises |
| 5. Capital and digital finance | Access modern capital markets and, where it fits, tokenized structures | Value trapped in an illiquid, opaque cap table |
The through-line is that the same asset powers all five stages: trusted, structured company data. Organizations with clean, integrated, well-governed data can scale AI pilots; those without this foundation struggle regardless of algorithmic sophistication. Stages 1 through 3 are the AI story. Stages 4 and 5 are where most executives stop thinking, and where the largest hidden value sits. This is the layer Stobox Intelligence is built for: the intelligence layer for companies preparing for the future economy, on the principle that AI is only as powerful as the quality of business information it can access.
The Second-Order Effect: The Data That Runs AI Is the Data Markets Price
The direct answer: the structured, verified data that makes AI work is now a driver of valuation and a precondition for raising capital.
Here is the point most AI coverage misses. The work of becoming AI-ready produces exactly the artifact that investors and acquirers now demand: clean, traceable, real-time business data. That is not a coincidence. It is the same underlying asset viewed from two directions.
Private markets have already priced this. The companies trading at premium multiples in 2026 aren’t just profitable. They’re AI-ready. The ones trading at discounts often carry what we call “AI debt,” structural liabilities invisible to traditional diligence frameworks that will cost the acquirer millions to remediate post-close. A company’s data infrastructure has become a line item in valuation. A target company’s relationship to artificial intelligence, its data infrastructure, workflow architecture, and organizational readiness for automation, has become a material factor in valuation.
Diligence itself now runs on AI, which raises the bar for the data being examined. The best tools promise 100% accuracy with every data point linked back to its raw source. A company whose numbers cannot be traced to source will not survive that scrutiny cleanly. The winning pattern in private equity is a single, structured, source-linked repository used from diligence through ownership: portfolio monitoring data is structured into investor-ready dashboards with full source traceability back to operational contracts and financial systems.
So the strategic case compounds. Build trusted data, and you get two returns from one investment: AI that actually works internally, and a company that is legible, valuable, and financeable externally. This is where the Stobox stack connects. The intelligence layer produces investor-ready data; Raisable is the infrastructure layer connecting investment-ready companies with modern capital markets; and Compass provides the tokenization infrastructure for compliant digital securities where a tokenized structure fits the asset. Intelligent, then investment-ready, then digitally connected, in that order.
How to Act on This
The direct answer depends on your seat, but the first move is the same for everyone: fix the data before you scale the agents.
If you are a CEO or founder. Stop counting licenses and pilots. Pick one high-value workflow, redesign it end-to-end with AI as the default, and instrument it so you can prove EBIT impact. Before that, audit whether your data is trustworthy enough to act on autonomously: is it consolidated, current, traceable, and governed. Treat this as one program that also produces investor-grade reporting, not two separate projects. A structured intelligence layer, such as Stobox Intelligence, is where that single source of truth lives. See the readiness view for how the stages sequence.
If you are an asset owner. Your data infrastructure is now part of your asset’s price. Weak, unverifiable data is a discount you pay at every raise and every exit. Building a verified, source-linked data foundation is the cheapest valuation improvement available to you, and it is the precondition for accessing modern capital or, where it fits, a tokenized structure via Compass.
If you are an investor. Add data and AI readiness to your diligence checklist as a material factor, not a footnote. Price “AI debt” explicitly. The companies with clean, governed, traceable data will scale AI, report faster, and clear diligence cleaner. Learn how these dynamics connect to capital formation in the /learn library or the /glossary.
The convergence is the whole point. The company that is intelligent is, by construction, closer to being investment-ready. The build is continuous.
FAQ
What is an AI-ready company? An AI-ready company is one whose business data is structured, verified, current, traceable, and governed to a standard where autonomous systems can safely act on it and external parties can trust it. It is defined by the state of the data and operating model, not by which AI model is used. Readiness, not model choice, is what separates high performers.
Why do so few companies see real value from AI if almost everyone uses it? Because adoption is easy and advantage is hard. In McKinsey’s 2025 survey, 88% reported using AI in at least one function, but only 6% qualified as high performers achieving more than 5% EBIT impact. The gap comes from failing to redesign workflows and from running AI on data that is not ready.
How does data readiness affect AI agents specifically? Agents act autonomously, so data errors become wrong actions rather than wrong numbers. When AI systems are put to productive use, shortcomings in data quality, governance, and interoperability can lead to operational failures. This significantly limits the potential for secure, large-scale automation using AI.
How many companies are actually ready to run AI agents in production? Few. Only 15 percent of companies are fully prepared to deploy agent-based AI in production, even though nearly 60 percent report investing tens or hundreds of millions in this area. Deployment is outrunning readiness.
Why are so many AI agent projects expected to fail? Mostly for non-technical reasons. Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value, or inadequate risk controls. The technology works; the organizations are not ready.
What is the single most important thing high performers do? They redesign the work, not just the tool. High performers were 3.6 times more likely to pursue enterprise-level organizational change, and 55% of them fundamentally redesigned workflows when deploying AI, compared to roughly 20% of other firms.
How does AI readiness affect company valuation? Directly. A target company’s data infrastructure, workflow architecture, and organizational readiness for automation has become a material factor in valuation. Companies with weak data carry “AI debt” and trade at a discount in 2026 diligence.
Can smaller companies compete on AI readiness against large enterprises? Yes, and often faster. Larger firms scale more, but size does not guarantee returns, and turnkey agentic tools have lowered the barrier for mid-market and SMB adopters. The decisive factor is trustworthy data and redesigned workflows, both of which are more achievable at smaller scale.
Where should a company start if its data is not ready? Start with the questions the AI will actually ask, not the entire data estate. Inventory the handful of sources a first workflow needs, verify and govern those, redesign that one process around AI, then measure impact against EBIT before expanding.
How does becoming AI-ready connect to raising capital? The trusted, source-linked data that makes AI work internally is the same data investors and acquirers demand. Building it once improves both AI outcomes and financeability, which is why the intelligent company is, by construction, closer to being investment-ready.
