AI's Agentic Work Units Misleadingly Measure Progress

Salesforce's Agentforce platform has reached $1.5 billion in annual recurring revenue, but behind the headlines, companies are struggling to translate activity volume into a stable production environment.

The problem lies in the measurement vacuum surrounding agentic AI, with five competing metrics that measure different things. Salesforce tracks Agentic Work Units, Gartner predicts 40% of projects will be canceled by 2027 due to costs and unclear value, McKinsey finds 93% of enterprises are overspending on AI, The Futurum Group notes a shift in ROI toward direct financial impact, and S&P Global Market Intelligence and MIT data show a gap between apps embedding AI and organizations running agents in production.

Anushree Verma, senior director analyst at Gartner, notes that most projects are still early-stage experiments driven by hype. This immaturity is distorting how companies spend their capital. The MIT NANDA report highlights that 95% of generative AI pilots fail to meet CFO expectations, underscoring the disconnect between initial excitement and actual business results.

The current misalignment in budget allocation is concerning. While back-office automation shows a measurable ROI for AI agents, more than 50% of GenAI budgets are being funneled into sales and marketing. Companies are chasing growth while ignoring operational efficiency that actually pays for the technology. McKinsey research shows that 60% of total agentic AI spend is consumed by iterative response refinement.

Keith Kirkpatrick of The Futurum Group explains that enterprises need to move toward standardized, cost-per-outcome metrics that account for the full lifecycle of an agent, not just the number of tasks it triggers. Decision-makers should ask harder questions: What is the cost of refinement cycles? How much of the budget is tied to actual margin improvement versus experimental overhead?