Business leaders rushing to deploy artificial intelligence must treat data readiness as a board-level imperative or risk generating costly “AI slop”, according to a senior executive at Capgemini.
In comments released to coincide with the World Economic Forum Annual Meeting in Davos, Kevin Campbell, CEO of Insights & Data at Capgemini, warned that while AI is advancing at breakneck speed, its success depends entirely on data integrity.
“Models trained on incomplete or inconsistent data don’t just make mistakes – they scale them,” Campbell said. “The result is expensive AI initiatives that fail to deliver measurable impact.”
He cautioned that without strict alignment between data quality and business outcomes, generative AI produces “outputs that look impressive but fail to create real value” — a phenomenon he termed “AI slop”.
The readiness gap
Despite the urgency, fewer than one in five organisations currently consider themselves data-ready, with most struggling to navigate integration, quality and governance issues.
Research cited by Campbell reveals that more than half of business leaders identify data quality and availability as major hurdles to accelerating AI adoption.
In response, 72 per cent of leaders say they will prioritise data foundations and pipelines over the next 12 months as the focus shifts from experimental pilots to enterprise-wide transformation.
Campbell argued that data readiness must move from being a “technical afterthought” to a strategic business metric, noting that missing just 5 per cent of data accuracy can derail entire initiatives.
“Inventory systems may show products available, yet none can be shipped. Vendor files might validate but lack banking details,” he explained. “Customer records appear clean but lead to billing errors and regulatory failures.”
Strategies for resilience
To bridge this gap, Capgemini outlined five strategic steps for organisations to build a robust foundation, starting with aligning data strategy with business outcomes such as customer experience and innovation.
Campbell also urged leaders to “invite AI to the table” by centralising data into a single source of truth for training, while implementing automated cleansing and anomaly detection.
Critically, he emphasised the need to proactively address bias by regularly auditing training data for fairness and representativeness.
“The future of AI is autonomous, adaptive and everywhere,” Campbell concluded. “But without robust, trusted data ecosystems, even the most advanced AI will fall short.”
