At a Glance

  • Provider: Walsh University
  • Credential: Applied AI: Building and Testing AI Workflows
  • Length: 8 contact hours (8 clock hours) in four 2-hour live virtual sessions with the instructor
  • Format: Live virtual, instructor led. Every session combines short technical explanations, demonstrations, individual practice, and hands-on work by participants.
  • Prerequisites: None. No programming is required.
  • Tools: An approved AI assistant such as Microsoft Copilot (or related company-specific platform) and a spreadsheet. Participants need a computer with a current browser and an account for the AI assistant, set up before the first session. The skills carry over to other AI tools.
  • Data: Participants use an approved AI tool and only information authorized for that environment. The instructor provides practice documents and data when workplace information cannot be used.

This course teaches professionals to build, test, and run an AI-assisted workflow. Participants extract structured data from documents with AI, configure a repeatable workflow with human review and exception handling, test and secure it, and demonstrate a working pilot with a one-page technical specification.

Learning Outcomes

By the end of this course, participants will be able to:

  • Write extraction instructions that return structured output, such as a defined table or schema, from unstructured documents.
  • Verify AI output against the source documents, require source references, and calculate field-level accuracy.
  • Configure a no-code workflow in an automation platform with a trigger, an AI step, structured output, a human approval step, and an output to a spreadsheet or record.
  • Add exception handling for missing fields and low-confidence results.
  • Build a test set of normal and edge cases, measure accuracy and consistency, and compare versions of the instructions.
  • Configure guardrails for input boundaries, sensitive data, human review thresholds, and logging, and map them to the four NIST AI RMF functions: Govern, Map, Measure, and Manage.
  • Document and demonstrate a pilot, including data flow, inputs, outputs, access, error handling, owner, and success measure.

Training Outline

Each 2-hour session is built from short technical explanations, instructor demonstrations, and hands-on work that participants complete during the session.

1. Working with Documents, Files, and Information (2 hours): 

  • Technical objectives: provide source documents as context and recognize the tool's access limits; write extraction instructions that return structured output (a defined table or schema) from unstructured documents; summarize, compare, and categorize across sources; identify missing or conflicting information, require references to source passages, and verify output against the source; calculate field-level accuracy.
  • Hands-on work in the session: the instructor demonstrates an extraction on one document, then participants process 20 provided documents (approved or fictional), compare sources, flag inconsistencies, extract 6 defined fields into a table, and score the results against an answer key.
  • Tools: Microsoft Copilot (or related company-specific platform).
  • Deliverable: a Document Analysis and Extraction Template and accuracy log.

2. From Prompts to Repeatable AI Workflows (2 hours):

  • Technical objectives: configure a no-code workflow in an automation platform with a trigger, defined inputs, an AI step with saved instructions, structured output, a human approval step, and an output to a spreadsheet or record; add exception handling for missing fields and low-confidence results; test with 10 inputs and refine the instructions from the results.
  • Hands-on work in the session: the instructor builds a first version live, then participants build their own workflow, run it on 10 inputs, and log every result.
  • Tools: Microsoft Copilot and Microsoft Power Automate (or related company-specific platform) and a spreadsheet.
  • Deliverable: a working AI-enabled workflow, workflow map, and test log.

3. Testing, Securing, and Monitoring AI Workflows (2 hours):

  • Technical objectives: build a test set of normal and edge cases; measure accuracy and consistency across repeated runs; version instructions and compare results; configure guardrails for input boundaries, handling of sensitive data, human review thresholds, and logging; record unresolved data, security, policy, and ownership questions; map each control to the four NIST AI RMF functions (Govern, Map, Measure, Manage).
  • Hands-on work in the session: the instructor demonstrates a test run and one guardrail, then participants run their Session 2 workflow against a 30-case test set, fix the three most frequent failures, and re-test.
  • Tools: the same tools and a spreadsheet for results.
  • Deliverable: a test report with before and after results.

4. Applying AI Tools to Technical Content and Workflow Tasks (2 hours):

  • Technical objectives: run the workflow as a pilot on a batch from start to finish; document data flow, inputs, outputs, access, and error handling; define an owner, a success measure, and a monitoring plan for a first experiment.
  • Hands-on work in the session: the instructor shows a sample one-page specification, then participants finalize and demonstrate the pilot, have a peer run it with new inputs, and record the results.
  • Tools: the same tools.
  • Deliverable: a working pilot and a one-page technical specification.

Assessment and Credentials

Each session ends with a deliverable that the instructor reviews. Walsh University awards the Applied AI: Building and Testing AI Workflows Certificate for attending all four sessions and completing the pilot.

Skills That Transfer Beyond

The skills taught in this course are not specific to any one employer. They are transferable across industries, because every organization using AI needs the same core abilities: pulling reliable structured information from documents, building repeatable workflows with human review, testing for accuracy, and controlling risk. The course teaches these methods with widely used AI and automation tools and practice documents, not any organization's systems, and maps its controls to the NIST AI Risk Management Framework (Govern, Map, Measure, Manage), a national standard used across sectors.

Data and Privacy in Class

Participants use an approved AI tool and only information authorized for that environment. The instructor provides practice documents and data when workplace information cannot be used.