FA-0596Agentic & Generative AIAutomation, RPA & Power Platform

Cassidy AI Essentials

Knowledge-grounded agents and a bounded business workflow

Create a scoped Cassidy agent and inspect a small knowledge-grounded workflow, with evidence checks and explicit human decision ownership.

Introduction

Why this course

This one-day workshop introduces Cassidy's Knowledge Base, Agents and Workflows through a selected business-document example. Participants practise clear instructions, approved knowledge selection, a small multi-step workflow and review of generated outputs.

An illustrative insurance-submission scenario shows extraction, completeness checking, guideline retrieval and summary drafting without transferring underwriting authority to AI. Exercises use synthetic or non-confidential documents and a learning workspace. Connected actions, approvals, sharing and usage limits depend on the workspace configuration; review steps can remain manual where native approval mechanisms are not available.

Learning outcomes

The course teaches participants to:

Explain the roles of Cassidy Knowledge Base, Agents, Workflows and chat.
Prepare approved documents and check source-grounded responses against the originals.
Write narrow agent instructions with inputs, output structure, boundaries and missing-information handling.
Create or refine a selected agent and small workflow using supported workspace features.
Apply human review to a synthetic document-intake or underwriting-support example.
Test outputs, access boundaries, failure handling and usage considerations before a pilot.

Prerequisites

Basic computer and web-browser skills
Familiarity with routine business documents and processes
Access to an approved Cassidy AI workspace
Sample or non-confidential training documents
No programming experience required

Training outline

6 modules

01Module 1 — Platform and Responsible Use5 topics

Knowledge Base, Agents, Workflows, chat and connected-app concepts.
Meeting-intelligence and integration features as availability-dependent overviews.
Plans, credits and usage; distinguish a feature demonstration from entitlement.
Hallucinations, privacy, approved sources, data minimisation and human accountability.
Check permissions and organisational restrictions before connecting data or tools.

02Module 2 — Knowledge and Chat4 topics

Organise selected documents in collections/folders and review import or sync options.
Compare available OneDrive or other connectors without assuming access inherits correctly in every configuration.
Choose appropriate knowledge, check conflicts/outdated material and verify cited responses.
Practise a business request with supporting context, output formatting, follow-up questions and source checking.

03Module 3 — Agent Instructions and Scope5 topics

Define the task, role, terminology, required inputs and output structure.
Specify rules, missing-information handling, escalation and quality-review criteria.
Create a narrow agent with supported natural-language setup and approved knowledge.
Review instructions, connector/read-write permissions, sharing and maintenance.
Test normal and boundary cases rather than assume grounding prevents errors.

04Module 4 — Workflow Prototype5 topics

Compare a conversational agent with a defined multi-step workflow.
Use a template, blank workflow or Workflow Copilot to draft a selected flow.
Review triggers, inputs, variables, AI/connected actions, conditional paths and structured outputs.
Insert explicit review/approval points: use supported native mechanisms where verified, otherwise a manual checkpoint.
Test one bounded learning run; keep record updates and external communications mocked or disabled.

05Module 5 — Illustrative Underwriting Support5 topics

Use synthetic submission documents for classification, extraction, completeness checks and guideline lookup.
Draft an evidence-linked risk-information summary, referral note or missing-information request.
Compare renewal or claims-history summaries using prepared sample data.
Review normalisation, exceptions, referral routing and reviewer decisions.
Underwriting, acceptance, pricing and coverage decisions remain with authorised professionals; the example is document support, not validated automated risk assessment.

06Module 6 — Testing and Pilot Handover5 topics

Define expected results and check sources, accuracy, consistency and edge cases.
Test access boundaries and failure handling; refine instructions and retain approved versions.
Review feedback, ongoing monitoring, credits and process ownership.
Prepare a small pilot outline with user responsibilities, documentation, change review and escalation.
Identify remaining organisational/legal review rather than claim compliance or production readiness.

A programme built around your team.

Share your training goals and requirements.