AI in Business: Tools & Techniques (MA-2042)

Learners practice the same task across multiple assistants to build judgment about which tool fits which job. They build their first custom assistant, configure a low-risk agent, run document and data analysis with multimodal input, and draft an AI acceptable-use policy for their team. Safety, privacy, and ethics are integrated into every module rather than reserved for a final block where they typically get cut for time.

The course is appropriate for any organization rolling out AI to non-technical staff, onboarding new employees to existing AI tools, or upskilling managers who need to direct AI-augmented teams.

Audience Profile

This course is designed for non-technical business professionals with zero to light AI exposure who need to use AI as part of their daily work. Typical roles include:

• Business managers and team leads
• Project coordinators and program managers
• Marketing, sales, and customer service professionals
• Operations, HR, and finance staff
• Executives and assistants supporting executives
• Any knowledge worker whose organization has rolled out Microsoft 365 Copilot, Google Gemini in Workspace, or a similar AI platform

At Course Completion

After completing this course, participants will be able to:

• Explain in plain language what a large language model is, what it can and cannot do, and why it sometimes hallucinates.
• Select the right assistant (ChatGPT, Gemini, Claude, Copilot, or Grok) for a given business task based on capability, data sensitivity, and licensing.
• Write prompts that produce usable first drafts for six common business artifacts: email, summary, outline, analysis, policy, and presentation.
• Use multimodal features — voice, image, document upload, screen sharing — for real work.
• Build a basic custom assistant (Custom GPT, Gemini Gem, Claude Project, or Copilot Agent) for a repeating task.
• Describe what an AI agent is, where agents fit in a workflow, and what risks they introduce.
• Apply their organization’s AI acceptable-use policy, or draft a starter policy if none exists.

Prerequisites

Comfort with a web browser and a work email account. No coding or technical background required. Learners should have access to at least one of the following before class:

• ChatGPT (free, Plus, or Enterprise)
• Google Gemini (free or via Google Workspace)
• Microsoft 365 Copilot (any licensed seat)
• ai (free or paid)
• Optional: Grok (X Premium)

Learners should bring a laptop with one supported browser installed. Microphone access is required for the voice-mode lab.

• Recognize and respond to the top five AI risks: data leakage, hallucination, bias, prompt injection, and copyright exposure.

Outline

What AI Actually is Today

Learning Objectives

• Define generative AI in plain language and distinguish it from traditional automation and machine learning.
• Identify the three model types every user encounters: fast chat, reasoning, and agentic.
• Explain why LLMs hallucinate and how context windows affect responses.

Topics

• Generative AI versus the broader AI umbrella
• How LLMs work — token prediction, the 90-second version
• Fast chat models, reasoning models, and agentic models
• Context windows, knowledge cutoffs, and recency limitations
• Why your assistant doesn’t know about something that happened last week

 The Five Assistants — Who’s Who and When to Use Which

Learning Objectives

• Compare the five major assistants on capability, ecosystem, and data handling.
• Apply a decision framework for selecting the right assistant for a given task.
• Run the same prompt across two assistants and evaluate the differences.

Topics

• ChatGPT (OpenAI) — Custom GPTs, voice mode, Deep Research, operator features
• Gemini (Google) — Workspace integration, large context window, Gems
• Claude (Anthropic) — writing and reasoning, Projects, Artifacts, Computer Use, MCP
• Microsoft Copilot — embedded in Word, Excel, Outlook, Teams; tenant-grounded
• Grok (xAI) — real-time X data, fewer guardrails, niche use
• Decision framework: data sensitivity, tenant grounding, capability, cost

Lab

• Lab 1: Run the same four-prompt exercise against two assistants of your choice. Compare outputs and discuss with a partner.

Prompting in 2026 — It’s Simpler Than You’ve Been Told

Learning Objectives

• Apply the four-part Role-Task-Context-Format prompting pattern.
• Choose between pasting, uploading, and linking content.
• Iteratively refine a prompt to improve output quality.

Topics

• Why traditional “prompt engineering” matters less than it did
• The Role, Task, Context, Format pattern
• When to paste, when to upload, when to link
• Iterative refinement and follow-up prompts
• System prompts, custom instructions, and memory features

Lab

• Lab 2: Rewrite three of your real work emails using each assistant’s tone and format controls.

Multimodal — Voice, Images, Documents, Video, Screens

Learning Objectives

• Use voice mode for thinking and drafting.
• Upload an image or document and extract structured information.
• Use screen-sharing or vision features to get help with on-screen tasks.

Topics

• Voice mode as a thinking partner
• Image understanding — screenshots, whiteboards, receipts, charts
• Document upload — PDFs, spreadsheets, slide decks
• Screen sharing and vision features in ChatGPT, Gemini Live, and Copilot Vision

Lab

• Lab 3: Dictate a meeting prep via voice. Upload the resulting document. Ask the assistant for a one-page executive summary.

 Productivity Inside the Tools You Already Own

Learning Objectives

• Use Microsoft 365 Copilot or Google Gemini inside Outlook, Word, Excel, and Teams.
• Decide when to use the embedded assistant versus the standalone chat app.
• Apply the “draft here, polish there” workflow.

Topics

• Microsoft 365 Copilot in Outlook, Word, Excel, Teams, PowerPoint
• Google Gemini in Gmail, Docs, Sheets, Meet
• Embedded versus standalone — when to use each
• The draft-here-polish-there workflow

Lab

• Lab 4: End-of-day email triage using Copilot or Gemini in your live inbox (or a sandboxed account).

Custom Assistants — Your First Reusable AI

Learning Objectives

• Explain when a repeating task justifies building a custom assistant.
• Build a working custom assistant on at least one platform.
• Decide whether to share, restrict, or keep an assistant private.

Topics

• Custom GPTs, Gemini Gems, Claude Projects, Copilot Agents — same idea, different platforms
• When to build one — any task you repeat weekly
• Anatomy: instructions, knowledge files, capabilities
• Sharing, governance, and audit considerations

Lab

• Lab 5: Build one working custom assistant for a real repeating task — meeting notes cleaner, proposal reviewer, or client research briefer.

 AI Agents — The 2026 Shift

Learning Objectives

• Distinguish between chat, copilot, and agent modes.
• Identify tasks suitable for agent delegation versus tasks requiring human judgment.
• Configure and supervise a low-risk agent task end to end.

Topics

• Definitions: chat (pull), copilot (assist), agent (act)
• What agents can do today — browse, fill forms, run code, send email, schedule, update records
• What agents still cannot do reliably
• Platforms: Copilot Studio agents, ChatGPT operator and tasks, Claude Computer Use and Agent SDK, Gemini agentic features
• Risks: runaway actions, prompt injection, credential exposure, audit trails

Lab

• Lab 6: Configure one low-risk agent task — calendar-to-email summary or document tagging. Watch it run. Discuss what you would and would not delegate.

Data Analysis and Decision Support

Learning Objectives

• Upload a spreadsheet and produce a useful analysis through natural-language prompts.
• Generate structured outputs including tables, charts, and decision matrices.
• Identify when an AI-generated number requires verification.

Topics

• Uploading spreadsheets — what works, what doesn’t
• Natural language to chart, pivot, and summary
• Structured outputs: tables, comparison matrices, decision frameworks
• Trust and verification — the “show me your work” prompt

Lab

• Lab 7: Analyze a provided 2,000-row sales dataset. Produce an executive summary. Spot the planted anomaly.

Safety, Privacy, and What Not to Paste

Learning Objectives

• Distinguish consumer and enterprise tiers of AI services.
• Identify the top five risks a business user is likely to encounter.
• Recognize warning signs that an AI response may be wrong or harmful.

Topics

• Consumer versus enterprise tiers — the most important distinction
• Data handling: training opt-outs, retention, region
• Top five risks: data leakage, hallucinated facts, biased outputs, prompt injection, copyright exposure
• Red flags — when your assistant is confidently wrong

Ethics, Policy, and Your Organization’s Rules

Learning Objectives

• Summarize the current AI regulatory landscape relevant to a US business user.
• Explain disclosure norms for AI-generated work.
• Draft a one-page AI acceptable-use policy for a small team.

Topics

• EU AI Act and US state-level AI regulations — 2026 snapshot
• Copyright, training data, and the practical fallout for business users
• Disclosure norms: when to tell colleagues, clients, and regulators
• Writing a one-page acceptable-use policy

Lab

• Lab 8: Draft your team’s AI do’s and don’ts in 15 minutes. Share two with the group.

Your 30-Day Plan

Learning Objectives

• Identify three habits that produce compounding returns from daily AI use.
• Commit to a specific 30-day implementation plan.
• Define personal success metrics for AI adoption.

Topics

• The three compounding habits
• Build one custom assistant. Run one agent task. Replace one meeting with voice-mode synthesis.
• Measurement: time saved per week, artifacts improved, decisions accelerated
• Resources for continued learning