Quail Group: Businesses fail at AI because of muddled data, not weak models

Most companies stall on AI not because of weak models or a lack of budget. According to the authors at Quail Group, the problem runs deeper: businesses do not understand which data really matters, so AI only helps them scale their own confusion faster.

It's not about the models

The market still expects that fresh investment will automatically deliver more intelligence and efficiency. The picture looks different. In an article for TNW, the authors cite State of Enterprise AI 2026: global AI spending may reach $2.52 trillion, yet only 14% of CFOs see measurable returns. Another warning sign: in 2025, 42% of companies abandoned most of their AI pilots. This looks less like isolated deployment mistakes and more like a systemic gap between ambition and actual execution.

The authors push back on the popular explanation that it all comes down to "dirty" data. Cleanliness matters, but on its own it does little if the data is not tied to specific decisions, not integrated across systems and not usable in day-to-day work. Many companies have piled up dashboards, reports and tracking systems that create an appearance of transparency. Yet teams often cannot explain why a metric changed, how it affects results, or what action should follow.

How the chaos grows

Scale makes the problem worse. Data volumes grow faster than companies' ability to interpret them. Teams measure everything they can, but do not always understand why. As a result, dozens of metrics compete for attention, definitions differ between departments, events are logged in different ways, and reporting often depends on manual corrections. In such an environment it is hard to build a single picture of the business: everyone works with fragments, and the fragments rarely match.

When AI is layered on top of this base, the ambiguity does not disappear. It spreads faster. Systems trained on contradictory inputs do not resolve ambiguity; they amplify it. According to figures cited in the piece, 61% of data leaders say that improving data quality helps move AI initiatives into production, but 50% still regard data quality and access as serious barriers. The gap between confidence and understanding is also worrying: 65% of executives believe employees trust data for AI, while 75% at the same time admit to gaps in data skills.

"AI and automation amplify the state of the data they rely on."

Where to start

The authors do not believe more convenient tools alone will fix this. If a company's internal processes are unclear from the outset, metric owners are undefined and the signals themselves are poorly described, any new AI system will run on top of the same organizational fog. So they propose starting not with new models or yet another dashboard, but with rebuilding decision-making logic. A practical start looks like this:

• Identify the questions the business finds hardest to answer today
• Assign owners to key data and metrics
• Standardize processes so events are logged the same way
• Cut redundant indicators and keep the signals tied to actions
• Build a coherent data layer that works for everyday use

The authors put particular emphasis on the human side. Even well-structured data will not help if the team does not know how to apply it in daily decisions. Change management is therefore not an optional add-on but part of the AI strategy itself. Companies need to teach people to tell meaningful signals from background noise and to act on them with confidence, rather than simply consume ever more reporting.

What this means

The article's main point is simple: AI does not cure organizational chaos automatically. If a company lacks clarity in processes, accountability and data, new models will only speed up the production of questionable conclusions. The winners will be those who first put their signals in order, and only then scale automation.