Analytics Legends The knowledge platform for SAP Analytics
Academy module

From PoC to Production — the SAP AI Delivery Lifecycle

From PoC to Production — the SAP AI Delivery Lifecycle — Analytics Legends section illustration for the SAP Analytics knowledge base (concepts, studies, Academy)

As of 2026-09-25

Edited by , sole editor and publication director — 27 years in the SAP ecosystem. How these pages are produced, and where automation is used.

Gives SAP AI delivery teams a lifecycle map from an approved pilot (M378) to a production system that survives model churn, drift and its first audit. Uses SAP's own AI Golden Path — Explore/Discover, Design, Deliver, Run & Scale, Evaluate & Improve, Extend — as the backbone, naming the SAP service that owns each phase (HANA Cloud PAL/APL/hana-ml for ML, the generative AI hub for single-turn generative work, Joule Studio and MCP-based agent tools for agentic work, AI Core resource groups for custom deployment). Covers the Design-phase decision record, Run & Scale as an evaluation-gated ramp rather than a single go-live, a continuous Evaluate & Improve cadence tied to SAP Note 3437766's documented model deprecations, and Extend as the test of reusability. Three exercises and a self-assessment gate the move to M361/M366/M367/M368.

What you will learn

  • Name SAP's own six-phase AI Golden Path (Explore/Discover, Design, Deliver, Run & Scale, Evaluate & Improve, Extend) and where SAP says most engineering effort belongs
  • Make the ML vs LLM vs agentic design decision explicitly, naming rejected alternatives and the SAP service each option would use
  • Distinguish deployment as SAP AI Core resource groups/scenarios/executables from Joule's package-and-feature activation flow
  • Plan Run & Scale as a ramp with evaluation-gated checkpoints instead of a single go-live date
  • Build a production evaluation cadence (Evaluate & Improve) tied explicitly to SAP's documented model-deprecation schedule
  • Explain what the Extend phase requires (documented interface, authorisation model, reusable configuration) for a solution to become a pattern rather than a one-off

Module overview

Who this is for. You have built a business case (M378) or inherited one; a sponsor has said yes to a pilot. This module is the map from that yes to a production system that survives its first model deprecation, its first audit and its first busy Monday. It uses SAP's own published delivery lifecycle — the AI Golden Path on SAP Architecture Center — as the backbone, names the SAP service that owns each phase, and marks the two gates most PoCs never pass: an evaluation harness before scale, and a change-management plan before rollout (M381). It assumes M333's fundamentals and feeds M361 (AI Launchpad operations), M366 (evaluation harness), M367 (LLMOps) and M368 (security for SAP AI).

Prerequisites

  • M333 (AI & LLM Fundamentals for SAP Consultants) and M378 (Building the SAP AI Business Case)
  • Working knowledge of at least one of SAP AI Core, the generative AI hub, or Joule Studio
  • Comfort reading a delivery plan or project phase gate

Outcomes

  • Map a real SAP AI initiative onto SAP's six AI Golden Path phases and identify which phases lack an owner or a deliverable.
  • Produce a written Design-phase decision record naming the chosen approach, the rejected alternatives, and the SAP service for each.
  • Design a Run & Scale ramp with evaluation-gated checkpoints instead of a single go-live milestone.
  • Specify an Evaluate & Improve cadence (test-set size, metrics, re-run frequency, alerting owner) tied to SAP's documented model-deprecation practice.

Full module available to members. The full module adds: the decision framework · the end-to-end scenario walkthrough · the KPI scorecard · the anti-patterns · the code blocks · the knowledge check · the diagrams.

Open in the app →

Next in this track