Prompting and AI Productivity for Office Work

A hands-on course for office, sales, operations and management staff who want to get real value from AI assistants such as ChatGPT, Claude, Gemini and Microsoft Copilot. You will learn how these tools behave, how to write strong prompts, iterate, use examples and templates, and apply AI to writing, summarizing, spreadsheets and planning. You will also learn to check outputs, protect data, handle bias and copyright, build a team prompt library, use custom assistants and built-in AI features, and measure time saved.

Course structure

• Module 1: Prompting Fundamentals (4 lessons). How AI assistants behave, the anatomy of a good prompt, iterating as a conversation, and examples and templates.
• Module 2: Everyday Tasks (4 lessons). Writing and editing, summarizing and extracting, spreadsheets and analysis, and brainstorming and planning.
• Module 3: Quality and Responsibility (4 lessons). Checking output, protecting data, bias, disclosure and copyright, and a team prompt library with guidelines.
• Module 4: Productivity at Scale (4 lessons). Custom assistants, AI inside everyday tools, measuring time saved and a 30-day plan.

Interactive labs

• Too vague, missing context, or strong?: Judge twelve prompts: is each too vague to get a useful answer, does it have a task but miss key context, or is it a strong prompt? Then check your answers.
• Which tasks should you try AI on first?: Place seven common office tasks on a grid of time spent against how well AI fits, to find the best first candidates.
• What AI saves on one task: Enter the weekly hours a task takes, the share AI saves after review time, your loaded hourly cost and tool costs to estimate hours freed, savings and payback.

How it works

• Work through 16 short lessons at your own pace. Each has key takeaways, an exercise for your own workplace and a quick knowledge check.
• Try the three interactive labs as you reach them in the course.
• Pass the 20-question final exam (70% or higher, unlimited retakes) to unlock your Chmltech certificate of completion.
• Free, with no sign-up. Your progress is saved in this browser.

AI chat assistants, such as ChatGPT, Claude, Gemini and Microsoft Copilot, are built on large language models. They are trained on huge amounts of text to predict likely next words, then tuned to follow instructions and hold a conversation. The result is a tool that can draft, summarize, rewrite, explain, brainstorm and analyze at remarkable speed.

What they are good at

• First drafts of emails, reports, proposals, job descriptions and procedures.
• Summarizing long documents, meeting notes and email threads.
• Rewriting for a different audience, tone or length.
• Extracting information into tables or lists.
• Brainstorming options and checking your thinking.
• Explaining unfamiliar topics, formulas or code.

Where they fall short

• Confident errors: they can state wrong facts, figures or references fluently, often called hallucinations.
• No knowledge of your business unless you provide it.
• Limited or dated knowledge of recent events, unless the tool can search the web.
• Arithmetic and precise counting can be unreliable unless the tool runs calculations or code.
• Varied output: the same prompt can give different answers each time.

The right mindset

Treat an AI assistant like a fast, capable new colleague who knows a lot in general but nothing about your company, and who sometimes makes things up. Give clear instructions and context, review the work, and stay responsible for what you send. Used this way, it saves real time on routine work. This course assumes your organization has approved AI tools; check your policy before using any tool with business information.

The quality of the answer depends mostly on the quality of the request. Vague prompts get generic answers. A good prompt gives the assistant what a human colleague would need.

Before and after

Weak: "Write an email to a customer about the wrong parts."

Strong: "You are a customer service manager at a 40-person industrial parts distributor. A long-time customer, a food processor, received the wrong parts twice this month, which delayed their maintenance. Draft a reply that apologizes sincerely, explains we've added a barcode check at packing, and offers a 10% credit on the two orders. Keep it under 150 words, professional and warm. Don't promise specific delivery dates."

The strong prompt takes 30 seconds longer to write and usually saves several rounds of editing. You don't need every element every time; context, task and format matter most. The lab after this lesson asks you to judge prompts.

The first answer is a draft, not the final product. The biggest productivity gains come from refining the output through a short conversation.

Useful follow-up moves

Let it ask questions

For bigger tasks, such as a proposal or a procedure, ask the assistant to list the information it needs before drafting. It will often raise points you had not thought of, and the draft will need fewer corrections.

Start fresh when needed

Long conversations can drift. If the assistant keeps repeating a mistake or mixing up details, start a new chat with a clean, complete prompt that includes what you have learned.

You stay the editor

Iterating is not about getting the AI to write perfectly; it is about getting a strong draft quickly that you then finish. Many people find the best workflow is AI for the first 80% and their own judgment for the last 20%.

Showing is often better than telling. When you want a specific style, structure or format, give the assistant one or more examples. This is sometimes called few-shot prompting.

When examples help most

• Matching your company's tone of voice in emails or social posts.
• Producing a consistent structure, such as a standard quote cover letter or incident report.
• Classifying or extracting information in a particular way.

Example: consistent product descriptions

"Write product descriptions in the same style as these two examples. Example 1: [text]. Example 2: [text]. Now write one for: stainless steel sanitary ball valve, 2-inch, tri-clamp ends, rated to 150 psi."

Build reusable templates

For tasks you do often, save a prompt template with placeholders:

"You are our sales coordinator. Write a quote cover email to [customer name] at [company] for [product/service]. Key points: [points]. Our lead time is [lead time]. Keep it under 120 words, friendly and professional, and end with a clear next step."

Fill in the brackets each time. Templates make results consistent across the team and save everyone from reinventing prompts. Later lessons show how to share them as a team prompt library.

Tables and structured output

Ask for output in a format you can use directly: "Return a table with columns Supplier, Item, Price, Lead time" or "Give me a numbered checklist". Structured output is easier to check and to paste into spreadsheets or documents.

Writing is the most common and often the most valuable use of AI assistants at work.

Common writing tasks

• Emails: replies to customers and suppliers, follow-ups, difficult messages, announcements.
• Documents: reports, proposals, procedures, policies, job descriptions, meeting agendas.
• Marketing: website text, social posts, newsletters, case studies.
• Editing: shortening, simplifying, fixing grammar, changing tone, translating.

Practical tips

• Start from your notes: paste your rough bullet points and ask for a polished draft. The content stays yours; the AI handles the wording.
• Specify the reader: "for a plant manager with little time" produces a very different email from "for our technical support team".
• Ask for plain language and short sentences when writing for customers or staff.
• Difficult messages: ask for two or three versions with different tones, then pick and blend.
• Translation: useful for drafts and understanding, but have important customer-facing or legal text checked by a fluent speaker.

Keep your voice

AI drafts can sound generic or overly formal. Give an example of your own writing and ask it to match your style, and edit the result so it sounds like you. Remove phrases you would never use. Customers notice when every email sounds the same.

Accuracy

Check every name, date, price, promise and technical detail before sending. The assistant may fill gaps with plausible but wrong details.

AI assistants are very good at condensing long material and pulling out specific information, provided you tell them what you need.

Summaries that are useful

• Say who the summary is for and what decision it supports: "Summarize this supplier proposal for our owner, who needs to decide whether to proceed. Focus on cost, schedule and risks."
• Ask for a structure: key points, decisions made, actions with owners, open questions.
• Set a length: "five bullet points" or "under 200 words".

Meeting notes

Paste a transcript or rough notes and ask: "List decisions, action items with owners and due dates, and unresolved issues." Many meeting and video-call tools now produce these summaries automatically. Always check that action items and owners are correct before circulating.

Extracting into tables

"From these five supplier quotes, extract supplier, item, unit price, minimum order, lead time and payment terms into a table." Extraction saves hours of re-typing, but verify a sample against the source documents, especially numbers.

Long documents

For contracts, standards or reports, ask targeted questions rather than for a general summary: "What are the termination clauses?" "What warranty does the supplier offer?" Ask the assistant to quote the section it relied on so you can check it. For legal or safety-critical documents, the summary is a starting point, not a substitute for reading the relevant sections or getting professional advice.

AI assistants can help with spreadsheets and analysis, but numbers need extra care.

Where AI helps

• Formulas: "Write an Excel formula that looks up the price in sheet Prices by part number in column A and returns 0 if not found."
• Explaining: "Explain what this formula does step by step."
• Cleaning data: "How do I split full names into first and last name columns?" or "Standardize these addresses."
• Analysis ideas: "What charts would best show monthly sales by region and product?"
• Uploaded files: some tools can analyze an uploaded spreadsheet by running code, producing totals, charts and summaries.

Checking numbers

• Language models are not calculators. When a tool does arithmetic in plain text, check it.
• Tools that run code to analyze data are more reliable, but check that they used the right columns, filters and date ranges.
• Spot-check results against a few rows you calculate yourself.
• Ask the assistant to show its working or the formula it used.

Example workflow

A sales manager uploads 12 months of order data and asks: "Show total sales by month and customer type, and flag customers whose orders fell more than 30% compared with the previous six months." The tool produces a table and chart in a minute. The manager checks totals against the accounting report and reviews the flagged customers before calling them.

Before uploading data, check your organization's rules: customer and financial data may need to stay within approved tools.

Beyond writing, AI assistants are useful partners for thinking: generating options, structuring plans and challenging your assumptions.

Brainstorming

• "Give me 15 ideas to reduce order-entry errors at a small distributor. Group them by effort."
• "Suggest names and taglines for our new maintenance service plan."
• "What questions should I ask in an interview for a production supervisor?"

Quantity first, judgment second: ask for many ideas, then pick and refine the best ones yourself.

Planning

• "Create a project plan for moving our warehouse to a new building in three months, with phases, key tasks and risks."
• "Turn these notes into a training plan for new customer service staff over their first two weeks."

Decision support

• "Compare leasing versus buying a forklift for a small warehouse. List the factors and questions I should answer."
• "Here is my plan. What are the three biggest risks, and what am I missing?"
• "Argue the opposite of my position as strongly as you can."

Asking the assistant to critique your plan or argue against it is one of the most valuable uses: it surfaces blind spots before you commit.

Keep decisions human

Use AI to widen your options and test your thinking, not to make decisions for you. It does not know your customers, team, finances or constraints unless you tell it, and even then it cannot weigh them as you can. The lab after this lesson helps you choose which tasks to try first.

Every AI output needs review before you use it. A simple, consistent check catches most problems.

A four-step check

1. Read it fully. Don't skim. Does it answer what you asked?
2. Verify facts and figures. Names, dates, prices, quantities, technical specifications, regulations and references. Check them against reliable sources or your own records.
3. Fix it. Correct errors, remove anything you can't verify, adjust the tone and add what's missing.
4. Own it. Once you send it, it is your work and your responsibility.

Typical errors to watch for

Higher stakes, more checking

A brainstorm list needs little checking. A customer quote, a safety procedure, a contract clause or anything about health, legal or financial matters needs careful verification, and often an expert's review. Match the depth of the check to the consequences of an error.

Ask for sources

Tools with web search can provide links. Open them: the source may not say what the summary claims. Without search, ask the assistant to flag anything it is unsure of, and treat unverified claims as unverified.

What you type into an AI tool may be stored, reviewed or used to improve the service, depending on the tool and your settings. Protecting business and personal information is essential.

Approved tools

Business and enterprise versions of AI tools usually offer stronger data protections, such as not using your data to train models, admin controls and data retention settings. Your organization should decide which tools are approved for which kinds of information. Use those, signed in with your work account.

A simple data rule

Practical habits

• Remove or replace names and identifying details when they aren't needed: "Customer A" instead of the real name.
• Don't paste whole databases or customer lists.
• Check retention and sharing settings, and be careful with shared chats and links.
• Ask when unsure.

Canadian privacy law, including PIPEDA, applies to personal information wherever it is processed, including in AI tools. When in doubt, leave it out. The Private and Local AI course covers options for keeping sensitive data in-house.

Responsible use goes beyond accuracy and data protection.

Bias and fairness

AI models learn from human-written text and can reflect its biases. This matters most in decisions about people. Be especially careful when using AI for:

• Screening résumés or writing job advertisements.
• Performance reviews or disciplinary matters.
• Credit, pricing or service decisions about customers.

Check outputs for stereotypes and unfair assumptions, keep a human responsible for decisions about people, and follow employment and human rights law.

Disclosure

Your organization should decide when AI use should be disclosed. Many companies don't require it for routine drafting that a person reviews, but do for content presented as personal or expert work, research or anything where the reader would reasonably expect to know. Never present AI-generated content as something it is not, such as a personal testimonial.

Copyright and ownership

• Don't ask AI to reproduce copyrighted material such as articles, books or song lyrics.
• Be cautious with AI-generated images and text for logos or key brand assets; ownership and originality questions are still developing.
• Check tool terms on how outputs may be used commercially.
• Respect confidentiality agreements: material shared with you under an NDA should not go into AI tools unless permitted.

Accountability

Whatever the tool produced, the person and organization using it remain accountable. "The AI wrote it" is not a defence for a wrong quote, a biased decision or a privacy breach.

Individual experiments become team productivity when good prompts and practices are shared.

Build a prompt library

Create a shared document or page with tested prompt templates for common tasks. For each, include:

• The task and when to use it.
• The prompt template with placeholders.
• An example of good output.
• Checks to do before using the output.
• The owner who maintains it.

Start with five to ten high-value prompts, such as customer reply drafts, meeting summaries, quote cover emails, job descriptions and procedure drafts. Add more as people find what works.

Simple team guidelines

• Which tools are approved, and for what information.
• The four-step check before using any output.
• Never share passwords, banking details or confidential data in unapproved tools.
• Humans stay responsible for decisions about people and for anything sent externally.
• Whom to ask with questions or to report a problem.

Learn together

Hold a short monthly session where team members share a prompt that saved time or a mistake they caught. Appoint one or two enthusiastic "AI champions" to help colleagues and keep the library current. Celebrate time saved, and be honest about tasks where AI did not help.

Most AI tools let you save standing instructions so you don't repeat the same context every time.

Options

• Custom instructions or memory: tell the tool about your role, company and preferred style once.
• Projects or workspaces: group chats with shared files and instructions, for example all work on one proposal or client.
• Custom assistants: many tools let you create an assistant for a specific job, with instructions and reference files, such as a "Quote Writer" that knows your products, terms and tone.

Designing a custom assistant

1. Pick one repeated task with clear inputs and outputs.
2. Write instructions: role, purpose, steps, tone, what to avoid, when to ask questions.
3. Add reference material: product sheets, price rules, examples of good output, style guide. Only include information approved for that tool.
4. Test with real cases and refine the instructions.
5. Share it with the team and name an owner to keep its files current.

Example

A distributor creates a "Customer Reply Assistant" with its returns policy, shipping terms, tone guide and five example replies. Staff paste a customer's email and get a draft reply that follows policy, which they review and send. Replies become faster and more consistent, and new staff learn the policies by seeing them applied.

Keep it current

A custom assistant with outdated price lists or policies gives confidently wrong answers. Review its reference files whenever policies change.

AI is increasingly built into software you already use: office suites, email, meeting tools, CRM, accounting and customer service platforms. These built-in features can save time without switching tools.

Common built-in features

Simple automations

Some tasks can be automated end to end with AI as one step, for example: incoming supplier invoices are read, key fields are extracted and a draft entry is created in accounting for a person to approve. Workflow tools connect apps and add AI steps without coding. Start with tasks that are frequent, rule-based and easy to check, and keep a human approval step for anything with financial or customer impact. The Workflow Automation course covers this in more depth.

Before turning features on

• Check the cost: some AI features are paid add-ons per user.
• Check data settings and where data is processed.
• Pilot with a small group and measure the time saved before rolling out to everyone.

To decide where to invest in AI tools and training, measure the benefit honestly.

A simple method

1. Pick a task, such as drafting customer replies or writing weekly reports.
2. Estimate how long it takes now, per instance and per week, across the people who do it.
3. Try AI on the task for two to four weeks, including the time spent reviewing and fixing output.
4. Estimate the new time per instance.
5. Note quality effects: fewer errors, faster responses, more consistency, or the reverse.

Value the time realistically

Time saved has value only if it is used: more customers served, faster quotes, less overtime, a hire avoided or time redirected to higher-value work. Don't count saved minutes as cash savings unless costs actually fall.

Example

A customer service team of four spends about 20 hours a week in total writing replies. With an assistant and templates, drafting time falls by about 40% after including review time, freeing roughly 8 hours a week, or over 400 hours a year. The team uses the time to call back customers with open issues, and response times improve. Tool licences cost a few hundred dollars a month.

Watch for hidden costs

• Time spent checking and correcting output.
• Errors that reach customers.
• Training and change effort.
• Licence costs for users who rarely use the tool.

The lab after this lesson calculates hours freed, savings and payback for one task.

Turn what you have learned into habits with a 30-day plan.

Daily habits

• Start routine writing with AI from your notes, then edit.
• Ask for critique on important drafts and plans.
• Check facts, figures and promises every time.
• Keep sensitive data out of unapproved tools.

Where to go next

The Large Language Models for Business Leaders course explains how these tools work in more depth and how to choose and govern them; AI Agents for Business Operations shows how AI can carry out multi-step tasks; and Private and Local AI for SMEs covers keeping sensitive work in-house.