Agent ROI and failure modes
Redesign the workflow around the agent, roughly 70% of value comes from that change, not the technology. Failures come from agentwashing, missing governance, and expecting the tool alone to deliver.
The 10-20-70 rule
About 10% of value comes from the algorithm, 20% from data and technology, and 70% from organizational and process change. Programs that invert this ratio and buy tools first are the ones that fail.
The 10-20-70 split is the single most useful predictor of agent success. Teams that spend their energy on the model and none on the workflow get an impressive demo and no ROI. Teams that redesign the process, scope the agent narrowly, and govern it capture the 70% that actually pays back. Vertical, task-specific agents make this concrete because each has a bounded job and a measurable success metric.
Frequently asked questions.
What is the most common cause of wasted agent spend?
Buying the technology without redesigning the workflow. Since ~70% of the value comes from process and organizational change, an agent dropped onto an unchanged process delivers a demo, not a return.