To provide assurance of trust and avoid compliance issues as well as to help mitigate risks, a ML system needs to be audited to assess compliance, reputational, legal, cyber and new product risks.
Try out the interactive web app https://processhorizon.com to design a best practice audit process for ML solutions
Nvidia's five-layer AI cake explains the industrial architecture for producing intelligence. The Event-driven SIPOC Action Architecture explains how that intelligence is transformed into governed human and AI actions, measurable outcomes and sustainable economic and societal value. I am thus proposing a further top layer representing enterprise value realization,
My objective was to provide one universal, intuitive and stakeholder-oriented Action Architecture for understanding, designing, governing and executing value-creating activities across the enterprise, from business modeling to autonomous AI operations. The architecture consists of five interconnected layers, each using the same Event-driven SIPOC methodology, with explicit objectives, trigger events, process
Governance defines what should be achieved. Risk Management determines what could go wrong. Compliance establishes what must be respected. Quality Management ensures that processes and their outcomes consistently fulfill requirements & intended objectives. Together, they establish an integrated management system for responsible human & AI-driven enterprise operations. Event-driven SIPOC model
My objective is for AI users to understand: how information becomes AI intelligence, how intelligence is used to perform tasks & how the entire journey is governed to produce reliable, safe & useful outcomes. The five-stage event-driven AI Operating & Control model follows the analogy of the Vacuum Cleaner sucking