Course Format and Delivery

Delivery Method: LiveOnline
Schedule: 2 sessions of 4.5 hours
Cost: $1,695 USD

All sessions are delivered live by an expert instructor in a fully interactive online environment.

*20% off for group bookings when booking 3 or more attendees from the same organization on the same course dates in the same transaction.

About this course

The Enterprise AI Engineering workshop is designed to equip participants with the knowledge and practical skills required to translate operational needs into effective AI solutions. This hands-on program introduces a forward-deployed engineering approach, emphasizing direct collaboration with stakeholders to understand business needs, assess data and systems, and select an appropriate solution. The workshop focuses on implementation decisions and customer outcomes rather than model development alone.

Participants will learn how to start with the operational problem, evaluate where AI can provide value, and move a clearly scoped pilot toward secure, maintainable production use. The workshop emphasizes solutions that users can trust, teams can support, and organizations can measure, helping participants connect technical implementation with practical business outcomes.

What you will learn

• Translate an organizational problem into a scoped AI use case with identified users, workflows, data sources, constraints, and success measures.
• Assess data readiness, enterprise integration requirements, and implementation dependencies.
• Distinguish when to use search, RAG, tool-using agents, agent workflows, or conventional automation.
• Identify security, privacy, governance, evaluation, release, and operational-support requirements.
• Define a bounded enterprise AI pilot with clear acceptance criteria, ownership, and human decision points.
• Create an AI Implementation Brief that supports stakeholder review and a pilot go/no-go decision.

Topics Covered

Module 1: Introduction to Forward-Deployed Engineering

• The course’s working definition of forward-deployed engineering: a customer-embedded approach connecting operational needs with implementation and delivery.
• Responsibilities across discovery, prototyping, integration, evaluation, deployment, adoption, and support.
• Working with sponsors, workflow owners, end users, IT, security, and data owners.
• Distinguishing a demonstration, proof of concept, pilot, and production service.
• Balancing rapid iteration with enterprise constraints and maintainability.
• Establishing decision ownership and avoiding tool-first projects.
• Hands-on exercise: Review an AI initiative that struggled to move beyond a demonstration. Identify gaps in user needs, data access, integration, evaluation, ownership, or adoption
  

Module 2: Customer Discovery and Stakeholder Alignment

• Identifying the sponsor, workflow owner, intended users, approvers, and support team.
• Separating the stated request from the underlying operational need.
• Understanding pain points, delays, repetitive work, inconsistent answers, and rework.
• Capturing budget, timeline, approved platforms, accuracy, latency, and risk constraints.
• Establishing a current-state baseline and measurable improvement.
• Defining stakeholder responsibilities and mandatory human decisions.
• Hands-on exercise: Activity: “Customer Discovery Interview.” Conduct a simulated stakeholder interview and turn a broad request into a use-case statement identifying the user, problem, constraints, and desired outcome.

Module 3: Workflow Definition and Pilot Scoping

• Mapping inputs, tasks, decisions, handoffs, outputs, and exceptions.
• Identifying escalation paths and consequences of incorrect or unavailable results.
• Determining which steps require AI and which can remain deterministic or manual.
• Defining pilot scope, exclusions, assumptions, and dependencies.
• Selecting a bounded improvement opportunity.
• Establishing initial acceptance criteria and adoption expectations.
• Hands-on exercise: Activity: “Workflow Discovery Map.” Map a current-state process and propose a bounded future-state workflow. Document what the pilot will do, what it will not do, and where human judgment remains required.

Module 4: Data, Systems, and Integration Readiness

• Inventorying document repositories, knowledge bases, databases, business applications, tickets, and APIs.
• Assessing source quality, freshness, completeness, duplication, contradictions, and authoritative status.
• Identifying source owners and update responsibilities.
• Reviewing data classification, permitted use, retention, and access restrictions.
• Preserving user-specific and document-level permissions.
• Assessing authentication, approved connectors, rate limits, network boundaries, and environment constraints.
• Defining synchronization, refresh, and failure-handling requirements.
• Preparing safe, representative development and evaluation data.
• Hands-on exercise: Activity: “Readiness Assessment.” Create a source-and-integration inventory. Classify prerequisites as ready for a pilot, requiring preparation or approval, or unavailable for the proposed scope.

Module 5: Selecting the AI Implementation Pattern

• Search for locating authoritative information.
• RAG for answering or summarizing using retrieved source material.
• Tool-using assistants for bounded interactions with approved systems.
• Defined workflows for predictable sequences with explicit decision points.
• Agentic workflows for dynamic planning or tool selection.
• Conventional automation for structured, deterministic processes.
• Comparing feasibility, cost, latency, reliability, integration effort, and support needs.
• Choosing read-only, draft-only, approval-required, or bounded-action behavior.
• Defining fallback behavior for unsupported answers or incomplete tasks.
• Hands-on exercise: Activity: “Pattern Selection and Architecture Sketch.” Compare candidate approaches for the same operational need. Select the simplest sufficient pattern and document its components, rationale, boundaries, and alternatives.

Module 6: Secure Architecture and Enterprise Governance

• Mapping the user interface, identity layer, model access, retrieval services, tools, enterprise systems, and logs.
• Identifying data flows, trust boundaries, and privileged operations.
• Applying least privilege to users, services, and tools.
• Separating read access from write or execution authority.
• Preserving source permissions through retrieval and response generation.
• Protecting credentials, sensitive data, and restricted business information.
• Assessing untrusted prompts, retrieved content, and tool outputs.
• Defining vendor data-handling, retention, and permitted-use requirements.
• Assigning ownership for risk acceptance, incidents, and service shutdown.
• Hands-on exercises: Activity: “Control and Ownership Map.” Annotate the proposed architecture with access restrictions, logging requirements, trust boundaries, control owners, and procedures for disabling the service or revoking access.

Module 7: Bounded Prototyping and Human Oversight

• Translating the selected pattern into a bounded prototype.
• Defining inputs, outputs, retrieved context, and permitted tool interactions.
• Establishing source traceability, citations, and unsupported-answer behavior.
• Validating model output before it drives downstream actions.
• Designing approval gates for communications, record changes, transactions, and deployments.
• Making reviewer responsibilities and exception handling explicit.
• Comparing a read-only assistant with an action-capable assistant.
• Hands-on exercises: Activity: “Prototype a Controlled Workflow.” Build or storyboard an internal support assistant using synthetic or approved sample data. Add a draft-for-review capability and identify the additional permissions, approvals, and controls it requires.

Module 8: Evaluation and Pilot Acceptance

• Building evaluation cases from representative user tasks.
• Testing normal, ambiguous, incomplete, unsupported, and adversarial requests.
• Evaluating correctness, source support, permission enforcement, and tool behavior.
• Measuring latency and cost against agreed constraints.
• Checking whether citations actually support the response.
• Separating technical performance from business value and user acceptance.
• Setting acceptance thresholds and release-blocking conditions before testing.
• Recording known limitations and unresolved risks.
• Hands-on exercises: Activity: “Pilot Evaluation and Go/No-Go Review.” Develop an evaluation set, assess sample outputs, and determine whether the pilot meets its acceptance criteria. Document blockers, required improvements, and approval owners.

Module 9: Deployment, Monitoring, and Operational Handoff

• Planning a controlled rollout with limited users, scoped data, and restricted capabilities.
• Defining escalation, fallback, rollback, and shutdown procedures.
• Monitoring errors, unsupported answers, tool failures, user feedback, and access violations.
• Planning source updates, model changes, and regression testing.
• Establishing runbooks, support ownership, and incident responsibilities.
• Preparing user training and adoption support.
• Measuring operational improvement against the original baseline.
• Defining the evidence required before expanding scope or autonomy.
• Hands-on exercises: Activity: “Production Readiness and Handoff Plan.” Create a release checklist and operational plan covering deployment owners, monitoring, support, rollback triggers, user training, and conditions for future expansion.

Module 10: Closing & Next Steps.

• Final exercise: Design an enterprise AI implementation plan for a real business or technical use case. Define the workflow, solution approach, security controls, evaluation criteria, and deployment plan. Present and discuss solutions.
• Key Takeaways and next steps for attendees.