A fresh global survey indicates that less than one in ten companies have the data quality and integration required to fully support AI at scale. In its latest quarterly "AI Momentum" report, business data firm Dun & Bradstreet found only 8% of businesses worldwide describe their enterprise data as “fully ready to support AI at scale” ([1]). While this is up from 6% in the previous quarter, it underscores that the vast majority of firms remain only partially or mostly ready ([2]). Crucially, the research shows a direct link between data readiness and results: 81% of the companies with fully AI-ready data report seeing strong or broad AI-generated returns on investment, compared to just 12% of those whose data is only “partially ready” ([3]). In other words, well-prepared data is translating into six times better AI business outcomes than poorly prepared data ([4]). This data readiness gap is a persistent theme, a similar study by IDC earlier this year found only 3.1% of organisations had achieved the highest level of AI maturity across strategy, technology, people and data governance ([5]). The slow progress suggests that many companies’ AI ambitions are still outpacing the quality and management of their underlying data.
The Dun & Bradstreet findings point to a growing recognition that AI success depends on data fundamentals. Companies continue to invest heavily in AI, 65% plan to increase AI spending in the next year, according to the survey ([6]), but this investment often runs ahead of efforts to organise and clean the data feeding AI models. As a result, the share of firms achieving broad or strong AI returns remains stuck at 28% overall, unchanged from last quarter ([7]). The report warns that enterprises may be “investing in AI faster than they are readying their data foundation to support it” ([8]). It recommends that executives double down on data quality, integration and stewardship if they want to turn experimental AI pilots into scalable, profitable business tools.
Industry and regulatory voices are amplifying the call to improve data governance and quality as prerequisites for AI at scale. In the banking sector, a Q3 survey by KPMG found 80% of senior technology executives in financial services identified data readiness and access issues as their top challenge in deploying AI agents ([1]). This was a sharp rise from earlier in the year and reflects growing awareness that AI initiatives will stall without reliable, well-governed data. Finance chiefs are increasingly taking ownership of AI data problems: Deloitte’s latest CFO Signals report found nearly 80% of large-company CFOs now play a formal role in setting AI investment frameworks and oversight ([2]). Many CFOs see unchecked AI spending and poor data quality as risks to the bottom line, citing the need for discipline in data management and cost governance ([3]).
New research on AI governance reinforces that data issues remain widespread even in advanced projects. An Ernst & Young (EY) survey of over 300 senior AI leaders, published in mid-September, revealed that in formal "AI assurance" audits conducted within the past year, data problems were the most commonly discovered issue ([4]). About 57% of these AI reviews uncovered data quality deficiencies, topping the list of concerns, ahead of issues like model drift or “shadow AI” (undocumented AI systems running without oversight) ([5]). Moreover, 72% of respondents in the EY study admitted they lacked the ability to fully trace and audit the data feeding their AI models ([6]), a worrying gap when regulators are increasingly focused on “evidence of data lineage and inputs” in high-stakes AI applications ([7]). The UK’s Financial Conduct Authority this month echoed these themes in new guidance for AI in finance, noting that using “carefully filtered high-quality datasets” and robust testing is vital to ensure AI systems produce fair and reliable outcomes ([8]). The clear message from auditors and regulators alike is that companies must tighten their data controls and quality checks as they scale up AI.
Even the most well-funded AI initiatives can be derailed by non-technical data issues, including compliance with environmental regulations. In a high-profile case late last week, a US federal judge allowed a lawsuit by the NAACP to proceed against Elon Musk’s new AI venture, xAI, over the operation of an AI supercomputing data centre in the US state of Mississippi ([1]) ([2]). The complaint alleges that xAI (also known as SpaceX AI or SpaceXAI, as it is owned by Musk’s aerospace company SpaceX) installed and ran 27 large gas-fuelled turbines to power its "Colossus" AI data centre without obtaining the required air quality permits ([3]). The unpermitted on-site power plant, built to support xAI’s energy-hungry AI systems, is accused of violating the US Clean Air Act by emitting harmful pollutants that threaten local communities ([4]). The judge’s Oct 31 decision to let the case move forward poses a serious potential setback for xAI’s infrastructure plans, and it highlights the growing climate and compliance challenges tied to AI’s data infrastructure.
The xAI case underscores that “AI readiness” is not merely a question of algorithms or IT architectures, but also of responsible data practices, regulatory compliance and public trust. In the rush to build ever-larger AI models and the data centres to power them, tech leaders are being reminded that environmental and governance standards still apply. Regulatory scrutiny is mounting: in October, the UK Information Commissioner’s Office extracted formal data protection pledges from ten major AI firms, including OpenAI, Google, Meta, and Microsoft, after a two-year review of how they handle personal data. And across many jurisdictions, from New Jersey townships to European courts, concerns over AI’s resource usage and data practices are triggering moratoriums, fines and legal battles. The lesson for enterprises is clear: neglecting compliance and ethical data management can quickly derail AI initiatives. Ensuring transparency, securing proper licenses, and engaging stakeholders early are becoming as integral to AI readiness as the technology itself.