AI Administrator

Design agentic workflows and practical automations that connect tools, information, decisions and human oversight — with no code.

A 28-hour, no-code training track for business administrators, operations managers, project coordinators, IT support and early-career professionals. Aligned with the IEEE CS Region 8 AI Caravan "AI Administrator" curriculum.

Modules

Hours

Hands-on labs

No code

Required

Course Modules

Fourteen two-hour sessions over seven days. Every module pairs the concepts with a guided, no-code lab you can do with free tools, and ends with a knowledge check in the style of the final exam. Three threads run through the whole course: workflow mapping, agents and automation, and controls with human-in-the-loop design.

Agents and Automation, Explained

What agents and automations are, the four parts of an agent, workflow versus agent, and real admin and operations examples.

Day 1 · Session 1

Prompting for Operators

Roles, context, constraints and structured outputs; reusable prompt templates; assistants with projects and uploaded files; spotting hallucination.

Day 1 · Session 2

Workflow Mapping I: The Current State

SIPOC, swimlanes in draw.io, touch time versus wait time, pain-point tags, and mapping interviews; a live AI map reviewer.

Day 2 · Session 3

Workflow Mapping II: Designing the Future State

Rule, model or person for every step; value and risk scoring; the prioritisation matrix and a one-line ROI; placing the approval gate; a live AI risk challenger.

Day 2 · Session 4

Automation Basics with n8n

Your first no-code automation: triggers, nodes and data items, IF branches and error paths, credentials, testing and publishing; three live recorded walkthroughs.

Day 3 · Session 5

Adding AI Steps to Automations

The four AI step types, where the step goes, a prompt with a fixed output, an 'unsure' route to a person and the cost per run; build an inbox triage in n8n with Gemini. Three live recorded walkthroughs.

Day 3 · Session 6

Connecting Tools, Data and Knowledge

Connectors, credentials, APIs and webhooks in plain words; MCP as the universal plug for agents; RAG for company documents and what may leave the building. Build a policy Q&A over Google Drive in n8n and see the same index used over MCP. Three live recorded walkthroughs.

Day 4 · Session 7

Building Your First Agent

Instructions, tools, memory and model in one AI Agent node; a five-part system prompt; tools that only read or draft; a ten-question trial shift with a prompt injection hidden in a calendar event. Build the Ops Desk assistant in n8n. Three live recorded walkthroughs.

Day 4 · Session 8

Multi-Step and Multi-Agent Workflows

Sequential, parallel and supervisor shapes; a writer and a critic with a hard cap, built and tested in n8n on three cases; hand-offs, MCP and A2A; when one prompt is enough; the always-on personal agents (Hermes, OpenClaw) shown live. Three live recorded walkthroughs.

Day 5 · Session 9

Controls I: Human-in-the-Loop Design

Five risk tiers, four human-in-the-loop patterns, interrupt and resume; an approval gate in front of every reply the triage flow sends, tested on three real e-mails: approved, rejected, timed out. Three live recorded walkthroughs.

Day 5 · Session 10

Controls II: Security, Privacy and Governance

Least privilege, prompt injection shown live on the Ops Desk agent and fixed in three layers, the ClawHub supply-chain case, data classes, NIST AI RMF, EU AI Act, ISO/IEC 42001 and a one-page AI policy. Three live recorded walkthroughs.

Day 6 · Session 11

Modules 1 to 11 are published. The remaining modules are released as the course progresses.

What you will be able to do

Map workflows

Draw a real business process as current-state and future-state maps, and score each step for automation value and risk.

Build agents and automations

Connect three or more tools in a no-code automation, add AI steps that classify, summarise and draft, and configure an agent with instructions, tools and memory.

Design controls

Tier actions by risk, add approval gates and audit trails, and apply security, privacy and governance rules so automation stays accountable to people.

Who it is for

Business administrators, operations managers, project coordinators, IT support specialists, students and early-career professionals who are comfortable with everyday workplace tools such as email, spreadsheets and chat. No programming or prior AI experience is required.

Tools you will use

Everything in the labs can be done on free tiers. Sign up before Day 1 so the first lab starts on time.

Purpose	Primary tool	Alternatives
AI assistant	ChatGPT or Claude	Microsoft Copilot, Google Gemini
Workflow mapping	draw.io (free)	Miro, Lucidchart, Visio
No-code automation	n8n (cloud trial or free self-hosted)	Make, Zapier, Power Automate
Agent builder	n8n AI Agent node	Copilot Studio, OpenAI Agent Builder, Dify, Flowise
Data and connectors	Google Workspace or Microsoft 365	Airtable, Notion, Slack, Teams
Always-on personal agents (demo only)	OpenClaw, Hermes Agent	ChatGPT agent mode, Claude with connectors
Governance references	NIST AI RMF, EU AI Act	ISO/IEC 42001

Assessment and certificate

On this site, mark every module complete to unlock your course certificate. In the instructor-led track, completion requires 80% attendance, a 70% score on the final exam (40 multiple-choice questions plus a timed practical exercise worth 40% of the grade), and a presented capstone project.

Certificate page Start Module 1