Agentic AI architecture
Bounded autonomy: an architecture contract for AI agents
A practical bounded-autonomy framework for AI agents: goals, scope, authority, budgets, invariants, stop conditions, and evidence before expansion.
In-person course
Agentic AI Architecture & Security
From tool-calling demos to bounded, authorized and resilient enterprise agents.
01 / The outcome
What you will be able to do
By the end of the course, you will be able to design and defend a bounded, authorized, observable and resilient agentic system with controlled business effects.
Learning outcomes
- Choose an appropriate agentic pattern and bounded autonomy level
- Design durable execution, state, retries and reconciliation
- Secure tools, MCP, identity, delegation and human approvals
- Control persistent memory and multi-agent collaboration
- Threat-model business effects and govern runtime and supply-chain risk
- Define goal-to-effect telemetry, evaluations, SLOs and incident response
- Defend the architecture with ADRs and explicit residual risk
02 / The experience
Practical architecture work, not passive theory
Two connected days built around one realistic enterprise case, ten architecture exercises and a production-architecture capstone. You will reason about execution, identity, tools, memory and operations without relying on a single framework.
Programme
- Choose the right agentic pattern, autonomy boundary and durable execution model
- Design the tool plane, MCP connections, identity, delegation and authorization
- Place human control around consequential effects and failure recovery
- Control memory and multi-agent coordination
- Threat-model the runtime, supply chain and business effects
- Define goal-to-effect telemetry, evaluations, incident response and residual risk
03 / What you take away
Included course materials
- A participant workbook with ten connected exercises
- Reusable technical examples and architecture decision structures
- Threat-model, control and evaluation artefacts for an agentic system
- A defensible production-architecture pack completed through the capstone
04 / Course fit
Is this course for you?
Check the intended audience and prerequisites before booking.
Designed for
- Solution, software, security, cloud, integration and platform architects
- Senior engineers and AI or agent platform engineers
Prerequisites
- Enterprise AI Architecture Bootcamp or equivalent understanding of RAG, enterprise identity, governance, evaluation and observability
- Basic familiarity with OAuth or OIDC, APIs, logging and architecture diagrams
05 / Related insights
Explore the architecture before the course
Start with a practical article drawn from the same teaching method.
Architecture review
AI architecture review: 12 questions before production
A practical 12-question AI architecture review checklist for business fit, data, RAG, agent authority, security, evaluation, operations, and ownership.
06 / Date and booking
Upcoming dates
Dates to be announced
Upcoming public dates will appear here once confirmed. Contact Anton to discuss the course, or request private training for your team at your location.