API Days Singapore 2026

FROM THEORY
TO THE BOARDROOM

Agentic AI Frameworks for Measurable Operational Impact

Andreas Kurz
Global Head of Digital Transformation, AlfaGomma Group
Adjunct Lecturer, NUS Business School
Hello!
AI
A2A MCP RAG Inference Token Foundation Model Orchestration Semantic Grounding Hyperautomation Prompt Injection Few-shot Zero-shot Multi-shot Retrieval-native Self-learning
01
Strategic
Human-to-AI-Agent
How to rethink and reengineer business processes from human workforce to AI agents. Starts with the executive mental model.
"Which processes can be translated into agentic AI workflows?"
02
Leadership
One Agent One Tool
Define and scope agent roles like employees. Create org charts and operating models for agent deployment.
"How can we integrate those agents into our existing organisation?"
03
Technical
3-Layer Agentic Architecture
Enterprise-ready agent harness for scalable and secure deployment of multi-agent systems.
"How can we ensure scalability, governance and security?"

Software programs that can understand the environment, make decisions and take actions based on context, rules and requirements provided without human intervention.

Objective
Tasks to solve, requirements
Capabilities
AI Model(s) connected
Context
Information, Rules, Policies
Tools
Access to other software programs, functions & APIs, permissions

Example: Document Automation

→
→
Copy link
→
Insert link
2 min
Per document
×
250
Per day
×
26
Production Plants
=
27 FTE
On Group Level
😵
💪 Get PDF
💪 Read PDF
💪 Classify Documents by page range
💪 Split into individual Documents
💪 Extract Document Number
💪 Validate if Document Number exists
💪 Archive Document
💪 Insert URL in System

How Organizations Scale with Separation of Concern

HR
IT
FC
OPS
Head of HR
Payroll
People Dev
Recruiting
ONE AGENT ONE TOOL
AGENTIC ORGANISATION DESIGN PRINCIPLE
Manager/Task Orchestrator
Receives task/request, plans & triages what to do, knows capabilities of team members
1 ⋮ n
Team Member/Workflow Executor
Receives curated tasks and context, plans based on context & available tool

Scaling of Agents & Enablement of A2A Communications while introducing robust guardrails & observability.

Hook Layer
Invocation
Message Queue Layer
Route
Manager / Orchestrator Queue
1
n Worker / Team Member Queue
Agent Layer
Execute
+
Objective
+
Model
+
Context
+
Tools
=
Next Action
FTE Savings
2
Full-time equivalents per Entity freed up from manual document processing
Human-to-AI Framework
Token Usage & Cost
↓ 60%
Specialist agents use smaller models, fewer tokens per task
One Agent One Tool
Time to Market
Months → Days
New agent from idea to production deployment
3-Layer Architecture