Redesign Your Workflows Before Adding More AI Agents

The article diagnoses a widespread problem: companies are treating AI as a tool to bolt onto existing workflows, when the real leverage lies in redesigning the workflows themselves. Research from McKinsey, BCG, Microsoft, and others cited in the piece converges on the same finding—organizations create measurable AI value only when they stop thinking in isolated use cases and instead map end-to-end processes, decide what humans should own and what agents should handle, and establish clear decision boundaries and review points.

A critical insight emerges from Johnson & Johnson's experience: even broad experimentation (nearly 900 GenAI use cases) does not automatically translate to business impact. The company's realization that 80% of value came from only 10% to 15% of initiatives suggests that most AI deployments are noise, not signal. The article argues this reflects a structural problem—teams layer AI on top of workflows that already suffer from manual handoffs, Excel copy-paste, and undocumented logic scattered across emails. Until those workflows are redesigned, AI becomes another layer of technical complexity on broken process foundations.

For executives, the implications are stark. BCG research shows nearly three-quarters of CEOs believe they are the main AI decision maker, and nearly all expect AI agents to produce measurable returns in 2026. Yet only 21% of organizations have a mature governance model for autonomous AI agents, and about 80% lack mature capabilities such as decision boundaries, real-time monitoring, and audit trails. This governance gap suggests that many organizations are measuring the wrong things (individual task performance) rather than the right things (workflow quality, business outcomes, and human accountability). The article frames this as a CEO-level operating discipline, not a technical problem to be solved by deploying more tools.