AI Productivity Starts With Redesigning the Work | Opinion

Global organizations are making a dangerous bet by restructuring their workforces around AI productivity gains without having the necessary operational models in place. Only a small fraction of companies have automated most of their processes, with legacy technology being the primary barrier to transformation. Companies are investing in AI capable of coordinating work across entire business processes, but they are still relying on structures built around departmental expertise.

This creates an expectation that AI will eliminate inefficiencies even though their operating models continue to create them.

Companies often make workforce decisions based on projected AI efficiencies that have not yet materialized. For example, in the order-to-cash process, delays, errors, and frustration can occur even when each department meets its performance targets. The root cause of this issue is that companies are organized around expertise rather than outcomes. Process Owners have been appointed for years to coordinate work across functional boundaries, but these leaders often lack the authority to make decisions.

The consequences of this mismatch between technological ambition and operational readiness can be severe. Companies may reduce headcount before the necessary processes and capabilities are in place, leading to failures and backlogs. Customers may experience delays, employees may inherit poorly designed workflows, and businesses may pay for technology that underdelivers.

The solution is not to reorganize the entire enterprise overnight. Instead, it involves identifying high-volume processes, such as purchase-to-pay, and appointing a leader accountable for their performance from beginning to end. This leader should have the authority to make decisions and report to various functional experts to resolve disputes.

By implementing AI and other process improvements in a focused manner, companies can achieve demonstrable improvements while retaining human oversight where necessary. This incremental approach addresses the issue of poor data foundations and allows for gradual improvements without waiting for a perfect enterprise-wide data environment.