AI project stalled? Blame your outdated, fragmented workflow - and redesign it now

A new ZDNET analysis argues that most enterprise AI projects stall not because of inadequate AI models, but because organizations layer AI onto outdated, fragmented workflows designed for a slower industrial era. The article cites McKinsey research showing 82% of supply chain leaders report new tariffs impacting 20-40% of their supply chain activity, creating volatility that manual processes can't handle. Meanwhile, Boston Consulting Group (BCG) research reveals an 'AI Adoption Puzzle' where nearly two-thirds of companies have moved beyond pilot projects, but only a small fraction successfully scale AI implementations.

The core problem identified is what the article calls the 'commercial operations bottleneck' - the manual connective layer between customer-facing systems and fulfillment/invoicing systems that relies on spreadsheets, email, and custom patches. This creates what author Vala Afshar describes as 'digital glue' where humans manually connect disparate systems, a workflow that collapses under modern tempo and uncertainty. The solution requires shifting from 'Systems of Record' (documenting history) to 'Systems of Agency' where agentic AI can reason, coordinate, and act across complex workflows.

Practical implications include redesigning organizational structures to allow AI agents to operate with trusted context and true agency, not just computational power. Companies winning what's termed the 'Cognitive Industrial Revolution' will be those that fundamentally redesign workflows rather than simply adding AI capabilities on top of legacy systems. This represents the third major revolution after steam power augmented human muscle and digital systems augmented human memory - now AI must augment human reasoning through properly designed agentic systems.