GLOBAL – October 27, 2026 -- Alteryx CEO Andy MacMillan said an AI system that recalculates revenue rules with every user query is not a strategy but an operational risk, as enterprises weigh the speed of generative analytics against the reliability of its outputs.
Alteryx outlines four-part VURA test for trusted AI answers
MacMillan described the company's VURA framework: logic must be visible, calculations understandable, outputs repeatable and results auditable back to source data. He said one clearly wrong or unverified answer erodes user trust almost instantly.
Golden Analytics CEO François Ajenstat said semantic models supply consistent definitions but do not make analytics accessible on their own. He placed the industry in Gartner's trough of disillusionment around semantic models and ungrounded AI. Golden Analytics supplies the exploration front end, while Alteryx supplies the governed workflows and business logic.
McGarrah Jessee CTO says traceable lineage drove internal adoption
McGarrah Jessee (McJ) CTO Bill Urciuoli said adoption moved once stakeholders could trace where a number originated and adjust the business question themselves. He named time already committed to legacy operating cadence, rather than technical debt or culture, as the biggest transformation barrier.
Analysts shift from Jira tickets to owning business questions
Urciuoli said McJ analysts moved from processing Jira tickets and static report builds to advising leadership on what to ask next. Ajenstat said AI amplifies analysts rather than replacing them, and MacMillan described them as "builders" who define the business rules behind enterprise-wide analysis.
MacMillan sets human accountability limit for autonomous agents
MacMillan separated trust in data, in an AI answer and in autonomous execution, and urged firms to assess the "blast radius" of automated decisions. AI can own background research, initial data preparation, pattern identification and routine workflow execution. Humans keep final decision ownership and risk assessment, even when an agent recommends or executes an action.