Technical Program Management for GenAI and Agentic Systems (GenAI TPM) (3 Half-Day Instructor Led)

GenAI and Agentic Systems for Technical Program Managers

Course offers 14 PMI PDUs upon successful completion, under ways of working

Get guidance from our team of educators

Our Community is Growing Every Day!

Course Overview

The Technical Program Management for GenAI and Agentic Systems bootcamp is intended for program
and delivery professionals who need to lead GenAI and agentic AI initiatives with confidence — without
getting trapped in vendor hype, shallow demos, or purely theoretical AI overviews. In this course, we
cover the full arc of an enterprise AI program: how modern GenAI systems actually work, how agentic
architectures differ from prompt-driven applications, how the emerging agent interoperability standards
(MCP, A2A, and Agent Skills) fit together, how to secure and govern agents that can take real actions,
and how to scope, sequence, measure, and scale AI initiatives that deliver productivity, operating
leverage, and revenue growth.

The bootcamp follows the TPM Institute philosophy: a practical blend of technical depth, program
management fundamentals, and hands-on exercises rather than slide-only theory. It is architecture-first
but TPM-oriented — you will not be asked to become a machine learning engineer, but you will leave
able to evaluate reference architectures, challenge oversimplified vendor claims, ask the right trade-off
questions, and lead AI initiatives across the major enterprise ecosystems from Google, Microsoft,
OpenAI, and Anthropic. At the end of this workshop, you will be able to lead the selection, design,
governance, and rollout of enterprise GenAI and agentic programs in a strategic, outcomes-oriented
manner.

Half-Day Sessions

We teach in three half-day sessions live over the internet to allow you the flexibility to fit your work and lifestyle commitments into the days you have class. This also gives you more time to spend, offline if needed, with the instructor.

Live Classes

Experience interactive, instructor-led sessions designed to enhance your learning through real-time discussions, live architecture whiteboarding, and problem-solving.

Hands-on Workshop

Interact with your fellow students to learn from their skills and experience while attacking the many design labs in the course. These exercises are built to fortify the skills you learned in lecture and will help you bring your new skills to your day-to-day job — including an opportunity heatmap, reference architecture sketches, an agent threat model, a platform fit matrix, and a 90-day rollout plan you can apply immediately.

Who Should Attend?

The bootcamp is intended to provide a perfect blend of AI technical knowledge and program management fundamentals. You should attend if you are a Technical Program Manager, Senior or Principal TPM, Program Manager, Project Manager, Agile Coach, Scrum Master, or a leader in the Project / Program Management Office who is moving into AI-enabled product, platform, or operational initiatives. If you are a technical delivery leader — such as an Engineering Manager, Architect, Technical Lead, or AI initiative lead — you should attend to strengthen your program delivery discipline, governance framing, and business-outcome orientation for GenAI and agentic programs. The course is equally valuable for TPMs supporting enterprise productivity, customer experience, knowledge management, analytics, automation, or revenue growth programs who must translate between architecture, engineering, product, security, legal, and operations.

What You Will Learn

How GenAI systems work end-to-end: foundation and reasoning models, tokens, context windows, embeddings, retrieval, grounding, tool use, memory, orchestration, and evaluation
How agentic systems work: planning, reasoning loops, tool invocation, multi-agent patterns, human-in-the-loop checkpoints, observability, and bounded autonomy
How the agent interoperability stack fits together: MCP for tools and data, A2A for agent-to-agent collaboration, and Agent Skills for portable procedural knowledge
How to secure agentic systems against the leading agentic threat categories, and why agents need first-class identity and lifecycle governance
How to compare the major enterprise agent platforms from Google, Microsoft, OpenAI, and Anthropic — and manage vendor churn and exit risk
How AI regulation and standards (including the EU AI Act, NIST AI RMF, and ISO/IEC 42001) shape program plans, stage gates, and documentation
How to scope the right first use cases, define a minimum viable feasible AI solution, sequence pilots to production, and measure productivity and revenue impact

Course Outline

A Course Outline is a comprehensive guide that details the key topics, learning objectives, teaching methods, and assessment strategies of a course. It serves as a roadmap for students, providing clarity on course expectations and ensuring a structured, engaging, and effective learning experience.

Section 1: The Technical Program Manager in the Agentic Era

Why TPM ownership matters more as organizations shift from GenAI pilots to agentic operations
Linking AI initiatives to productivity, cost, quality, and revenue outcomes
The TPM triangle applied to GenAI → technologist, agilist, and program leader
The enterprise AI arc → chatbots → RAG → agents → interoperable multi-agent systems
Team Exercise – AI opportunity heatmap — map 6–8 candidate AI opportunities in your organization by value, technical feasibility, and change-management complexity

Section 2: How GenAI Actually Works

Foundation models, tokens, prompts, context windows, and inference
Reasoning models and extended thinking → capability, latency, and cost trade-offs
Embeddings, semantic search, retrieval-augmented generation (RAG), and grounding
Context engineering → system prompts, retrieved context, memory, and compaction for long-running work
Latency, cost, hallucination, drift, and evaluation basics
When fine-tuning is not the answer
Team Exercise – Architecture teardown — identify the difference between a plain prompt workflow, a grounded RAG workflow, a workflow with tools and memory, and a reasoning-model workflow

Section 3: Enterprise GenAI Architecture & Use-Case Selection

Reference architecture → user/channel, orchestration layer, models, tools, memory, data sources, observability, and guardrails
Agent-ready data → quality, permissions-aware retrieval, and knowledge lifecycle
Security, privacy, identity, data boundaries, and approval patterns
Build-vs-buy considerations and integration realities
Evaluation-first design → defining evals before building
Minimum viable feasible product for GenAI programs
Use-case selection criteria → frequency, friction, data availability, human-review requirements, and adoption likelihood
Business case framing for productivity and revenue use cases, and sequencing pilot to production
Team Exercise – Draw a reference architecture for an internal knowledge assistant and identify the top five architecture and governance risks
Team Exercise – Use-case scoring canvas — rank candidate use cases such as knowledge assistance, proposal acceleration, account research, service triage, or engineering support

Section 4: Agentic Systems & Agent Architecture

Prompted responses vs. workflows vs. agents vs. multi-agent systems — and when not to use agents
Planning, tool use, state, memory, and action loops
Least agency → granting the minimum autonomy required for safe, bounded tasks
Agent loop components → goals, instructions, tools, policies, memory, environment, and evaluator
Human-in-the-loop design and approval gates
Observability, traces, evaluation, and rollback
Failure modes → runaway loops, over-delegation, unsafe actions, weak grounding, and cascading failures across connected agents
Team Exercise – Continuum mapping — classify example solutions as chat, RAG, workflow automation, single-agent, or multi-agent, and assign each an appropriate autonomy level
Team Exercise – Whiteboard an agent for sales account planning or incident triage, including tools, approvals, autonomy boundaries, and fallback paths

Section 5: The Agent Interoperability Stack — MCP, A2A & Agent Skills

Model Context Protocol (MCP) architecture → clients, servers, tools, resources, prompts, and security boundaries
Agent2Agent (A2A) protocol → agent cards, task lifecycle, and cross-vendor agent collaboration under open governance
Agent Skills → packaging procedural knowledge as portable, progressively disclosed capabilities
How the three layers complement each other in a single enterprise architecture
Where emerging agentic-commerce protocols fit, at a briefing level
Governance → vetting MCP servers, skills, and remote agents as supply-chain components
Team Exercise – Layer mapping — for one enterprise workflow, identify what belongs in MCP tools, what belongs in a skill, and where agent-to-agent hand-offs occur

Section 6: The Enterprise Agent Platform Landscape

Google Gemini Enterprise Agent Platform → Agent Studio, Agent Development Kit (ADK), Agent Engine managed runtime, Model Garden, sessions and memory, and governance
Microsoft → Foundry Agent Service, Agent 365 as the enterprise agent control plane, and Entra Agent ID for agent identity
OpenAI → AgentKit, Agents SDK, ChatKit, Connector Registry, and evals — including a live case study in managing vendor deprecation risk
Anthropic → the Claude Agent SDK, subagents, permissions, hooks, and Skills
Platform selection criteria → existing cloud estate, identity integration, protocol support, exit costs, and roadmap stability
Program controls for agentic delivery → architecture review checkpoints, safety review, legal and compliance touchpoints, pilot metrics, and release governance
Team Exercise – Platform fit matrix — compare the ecosystems for an internal enterprise assistant and a cross-functional workflow automation scenario, including a deprecation and exit-risk column
Team Exercise – Draft a stage-gate checklist for an agentic pilot moving toward production

Section 7: Agent Security, Identity, Governance & Regulation

The leading agentic threat categories → agent goal hijack, indirect prompt injection, memory poisoning, tool and supply-chain compromise, cascading failures, and rogue agents
Why agents need first-class identity → authentication, least-privilege access, conditional access, sponsorship, and lifecycle management
Managing agent sprawl → inventory, registry, ownership, and kill-switch and containment patterns
Decision rights across product, architecture, security, legal, data, and operations — what the TPM owns vs. what security owns
AI regulation as a program input → the EU AI Act phased timeline, high-risk classification, and transparency obligations; NIST AI RMF and ISO/IEC 42001 as management-system scaffolding
Model risk, privacy, content safety, vendor management, change management, enablement, and adoption
Team Exercise – Threat-model one agent design from earlier in the course against the leading agentic threat categories and define its identity, access, and containment controls
Team Exercise – Build a RACI for a GenAI or agentic initiative involving business, IT, security, legal, and data owners — including an AI-inventory and risk-classification step

Section 8: The Agent Interoperability Stack — MCP, A2A & Agent Skills

How agent harnesses work → repo and file understanding, edits, command execution, approvals, and sandboxing
Subagents, skills, and long-running background work
Computer-use and browser agents → capabilities, current limits, and risk posture
Where code and computer-use agents fit in the SDLC and knowledge work; evaluating developer-productivity claims
Productivity metrics → cycle time, throughput, rework, service levels, and knowledge reuse
Revenue metrics → conversion support, account expansion enablement, speed to proposal, support efficiency, and retention signals
Leading vs. lagging indicators, and how to avoid vanity metrics
Portfolio balancing across quick wins, strategic bets, and foundational enablers; budgeting for pilots vs. platform build-out; architecture debt and reuse
Team Exercise – Compare a general-purpose enterprise agent architecture with a code-centric engineering workflow and define distinct controls and KPIs for each
Team Exercise – Metrics design lab — define a KPI set for one productivity use case and one revenue-adjacent use case, then place both on a 12-month roadmap

Section 9: Capstone Simulation & Executive Readout

Teams synthesize architecture, protocol choices, security controls, governance, use-case priority, metrics, and rollout plan into an executive recommendation
Team Exercise – Capstone — present a 90-day TPM plan for launching an enterprise GenAI or agentic initiative, including identity, security, and compliance checkpoints

What You Will Leave With

AI opportunity heatmap for your business unit or program area
Reference architecture sketch for a GenAI assistant or agentic workflow, annotated with MCP, A2A, and Skills layers
Use-case scoring sheet with feasibility, value, risk, and adoption dimensions
Agent threat-model worksheet with identity and containment controls
Pilot stage-gate checklist with architecture, security, identity, legal/regulatory, and adoption checkpoints
Platform fit matrix including protocol support and vendor exit-risk scoring
Metrics framework distinguishing productivity, quality, risk, and revenue indicators
90-day rollout plan for one candidate GenAI or agentic initiative

Get Started

Start Your Course Today!

Unlock your potential with TPM Institute — empowering technical program managers to lead the agentic era. Start your journey today!

Trusted By Multiple Fortune 500 Companies.