Artificial intelligence is moving beyond pilots and proofs of concept into what leaders describe as “invisible infrastructure.” Enterprises must rethink operating models with AI embedded directly into decision-making, governance, and workflows rather than treating it as a bolt-on tool.
Kelvin Cheema, Global CIO and Managing Director, Global Transformation & Change at Acuity Analytics, said AI delivers limited value when confined to analytics sandboxes or isolated automation projects. Instead, AI becomes transformational when it influences core processes such as financial close, procurement, forecasting, and risk management.
Cheema described this shift as moving toward “enterprise as code,” where decisions, processes, and governance are structured, testable, and adaptive. In this model, AI is inseparable from how the organisation operates.
However, many AI projects fail. Cheema noted that organisations often start by searching for use cases rather than redesigning end-to-end workflows. Without clear business ownership, defined value metrics, and unified data, pilots stall. He estimates fewer than 5% of organisations are truly “AI future-ready,” meaning AI is scaled into measurable business outcomes.
Key barriers are organisational, not technical. Siloed data, fragmented systems, and weak governance undermine even advanced models. Cheema emphasised that integration is critical: “Layering AI on fragmented data leads to biased outputs and scale stagnation. Integration beats raw intelligence.”
At Acuity Analytics, consolidation into a cloud-based enterprise stack—including ERP, HCM, and unified data architecture—created a governed foundation for AI deployment.
Success metrics extend beyond adoption rates or cost savings. Enterprises should track improvements in forecast accuracy, decision quality, cycle times, revenue impact, and governance maturity, including explainability and audit trails.
Cheema predicts that over the next three to five years, competitive advantage will depend less on algorithms and more on operating model design, governance maturity, and effective human-AI collaboration.
Source:
