ServiceNow thinks you're doing AI fast and wrong (www.theregister.com)

ServiceNow has started rolling out an "AI Experience" — a set of agentic, multimodal features aimed at letting users drive ServiceNow workflows by voice and by letting AI read screenshots/forms to kick off processes. The company rushed this outside its usual biannual cadence so customers could access AI capabilities sooner; the current shipping piece is "AI Lens" (automatically interpreting images/forms to instantiate workflows), while the headline "voice agents" and the back-end plumbing for autonomous workflow actions are demoed now but promised to arrive by end of 2025. ServiceNow customers run private instances and can opt into upgrades on their own schedule. The move matters because it signals enterprise software shifting from augmentation to agentic automation — NLP-driven voice agents, multimodal input, automated ticket creation and recommended fixes imply tighter integration of root-cause analysis, change actions and ITSM workflows. That promises headcount and time savings but increases vendor lock-in, requires robust security/audit trails, and raises governance questions about autonomy and error handling. ServiceNow’s CIO warns companies are applying AI to old tasks (RPA 2.0) and urges outcome-oriented prompts (e.g., “summarize and advise how to improve”), underscoring a push toward higher-level, outcome-driven agent design rather than one-to-one task automation.