The Business Processes Best Suited to AI-Driven Automation

The processes best suited to AI-driven automation share three traits. The work repeats at high volume, individual cases vary too much for predefined rules, and checking the AI’s output stays affordable as volume grows.

When one of those traits is missing, rules-based automation or people usually do the job more cheaply.

At SAP Connect 2026, CEO Christian Klein said customers can set agent autonomy process by process, keeping human approval where risk is high and automating more routine work. That leaves each enterprise to decide which processes qualify.

Variation Separates AI Candidates From Rules-Based Work

Rules-based automation handles steps with known inputs and outcomes. AI earns its cost where a repeatable process keeps producing cases that differ in small ways. Examples include a payment without a usable invoice reference, a supplier email that mixes a price query with a delivery change, or a dispute that requires checking several records. The pattern is familiar, yet no fixed rule covers every version. Gartner drew a similar distinction in 2025, advising agents where decisions are needed and conventional automation for routine workflows.

Volume and Review Costs Decide the Economics

Agents carry fixed costs for infrastructure, orchestration, and the specialists who keep them running. McKinsey’s analysis of agentic workflow economics shows volume absorbing those costs. An onboarding agent serving 2,500 customers a year cost $10,000 to $15,000 to run, and doubling the volume raised that only to $15,000 to $20,000.

Review is the bigger variable. In a McKinsey banking customer service example, human oversight made up about 70% to 75% of an agent’s variable running costs, compared with 20% to 25% for tokens. The strongest candidates send only low-confidence or high-value cases to experts. Low-volume, high-stakes decisions sit at the other extreme, since nearly every output needs expert review and the fixed costs never spread.

SAP Users See Early Value in Routing and Service Work

SAPinsider’s June 2026 research found workflow automation and task routing to be the use case most often delivering value, at 30%, ahead of conversational interfaces at 28% and decision support at 24%. Customer service and support led the areas where respondents have embedded AI, at 30%. IT operations and Basis, order-to-cash, procure-to-pay, and finance followed at 23% each. All of these are high-volume functions with steady streams of similar but nonidentical cases.

Starting conditions also shape results. SAPinsider’s Robert Holland found that customers reporting progress at SAP Connect had invested in foundations such as process standardization. They also began with a defined business problem before deciding where AI fits.

What This Means for SAPinsiders

Score processes on variation, volume, and review cost. Assess all three before choosing among rules, AI, or people. Spending then goes where AI can recover its fixed costs.

Start with standardized, high-volume work. Routing, matching, and service queues in harmonized processes give agents consistent context and enough volume to spread costs. Teams see value sooner, with fewer surprises in production.

Tie autonomy to process risk. Route only low-confidence or high-value cases to experts. Oversight costs then stay contained as volume grows.