On this page
Most people’s first honest experience with AI at work goes like this: they type a short question into ChatGPT or Microsoft Copilot, they get a generic answer, and they conclude that AI is overhyped. They go back to doing the work the hard way.
The tool isn’t the problem. The prompt is.
A generic prompt produces a generic answer. A structured prompt produces something you can actually use. And the gap between the two is smaller than you’d think — it comes down to four parts that fit on a sticky note.
This post walks through the framework we teach to client teams across Northern Ontario when we roll out Microsoft 365 Copilot. It works for ChatGPT, Copilot, Gemini, Claude, and any other modern AI tool. No technical background required.
Think of AI Like a Very Fast New Hire
If you hired a sharp but brand-new junior staff member and said “write me a job description,” they’d produce something generic. Not because they’re incapable — because you didn’t tell them what the role was, who it was for, what to reference, or what good looked like.
AI is the same. The quality of the output scales directly with the quality of the brief.
The GCSE Framework: Goal, Context, Source, Expectations
Four slots. Fill them every time. That’s the whole framework.
1. Goal — What You Want
The outcome, stated clearly and specifically. Not “help me with X” — “produce Y.”
- Weak: “Write something about our new service.”
- Strong: “Write a 150-word LinkedIn post announcing our new managed backup service.”
The difference is not length. The difference is specificity.
2. Context — Why It Matters and Who It’s For
The background the AI needs to make useful decisions. Audience, purpose, constraints, tone, any important history.
- Weak: “Write a follow-up email.”
- Strong: “Write a follow-up email to the owner of a 15-person construction firm in Thunder Bay. I met him at a trade show last week. He’s evaluating three IT providers, including us. Tone should be professional but not stiff — we’re a small local shop, not a corporate vendor.”
This is where most prompts fail. People assume the AI knows things it cannot possibly know.
3. Source — What the AI Should Draw From
Which documents, data, or examples should inform the answer. For Microsoft Copilot and other file-aware tools, this can be an actual attached file, a SharePoint link, or pasted text. For general chat, it’s text you provide in the prompt.
- Weak: “Summarize our services.”
- Strong: “Summarize our services based on the attached one-pager and the FAQ section of our website (copied below).”
If you don’t give the AI a source, it makes one up. That’s not useful. It’s dangerous.
4. Expectations — What the Output Should Look Like
Format, length, tone, structure, and explicit do/don’t rules. This is the slot that separates “draft I have to rewrite” from “draft I can send.”
- Weak: “Write a proposal.”
- Strong: “Write a two-page proposal with a one-paragraph executive summary, a bulleted scope list, a three-row pricing table, and a call-to-action at the end. Plain language, no jargon, no em-dashes. Do not include phrases like ‘leverage synergies’ or ‘best-in-class.’”
Telling AI what not to do is often more powerful than telling it what to do. Negative instructions tighten output faster than five rounds of “make it better.”
Good → Better → Best: One Prompt, Three Versions
Same task — prepare for a candidate interview for a new office administrator.
Good:
“Help me prepare interview questions for an office administrator role.”
Result: a generic list you could have pulled from any HR blog.
Better:
“I’m the owner of a managed IT company in Thunder Bay hiring a new office administrator. Please generate interview questions that cover experience, culture fit, and attention to detail.”
Result: usable, but still not tailored to the actual candidate.
Best:
“I’m the owner of a managed IT company in Thunder Bay hiring a new office administrator. Review the attached candidate resume and the attached job description. Generate 12 interview questions: 6 experience-based (tied to specific projects on the resume), 3 scenario-based (how they’d handle a typical day in our office — client calls, ticketing system, scheduling), and 3 culture-fit. For each question, add a 1-line note on what a strong answer would include. Do not use generic HR questions.”
Result: a brief you can walk into the interview with.
The best version isn’t harder to write. It just has the four parts.
Extend the Prompt’s Purpose
Before you write the prompt, ask yourself: “What am I actually going to do with the output?”
If you’re going to send it in an email, ask the AI to write the email — not just the content of it. If you’re going to turn it into a slide deck, ask for slide-ready structure. If it’s going into a spreadsheet, ask for a table.
Example:
- Prompt A: “Recap yesterday’s client meeting with action items.”
- Prompt B: “Write a follow-up email to the attendees of yesterday’s client meeting. Include a short thank-you opener, a table of decisions made, a second table of action items with owners and due dates, and a short list of agenda items for our next meeting.”
Prompt B skips a step. That’s where the real productivity gains live.
Keep the Conversation Going
AI is iterative, not one-shot. The first response is a draft, not a deliverable. The people who get the most out of AI build this habit:
- First pass — get the initial draft
- Narrow — “Focus only on the construction industry examples.”
- Reformat — “Turn the second section into a bulleted list.”
- Sharpen — “Rewrite the intro with a stronger hook. Lead with the problem, not the company.”
- Check — “What did I miss that a client might push back on?”
Most people stop at step 1 and then complain about the output. The ones who treat AI as a conversation get results that rival what they’d produce themselves — in a fraction of the time.
A Real Workflow: Turning a Meeting Into a Proposal in 15 Minutes
Here’s how a sales lead at a DVG client uses this framework weekly. The meeting is with a prospective client. Transcription is enabled in Microsoft Teams. After the meeting:
Prompt 1 (Goal + Source):
“Based on the attached Teams meeting transcript, extract the prospect’s three top pain points, the services we discussed, and any timeline or budget cues they dropped.”
Prompt 2 (Goal + Context + Expectations, using the output of Prompt 1):
“Using the pain points above, write a two-page proposal tailored to this prospect. Open with a one-paragraph restatement of their challenges in their own language. Include a scope list, a three-row pricing table ($X/$Y/$Z tiers), and a call-to-action. Plain language, no buzzwords, Canadian English.”
Prompt 3 (Review):
“Pretend you are the prospect. Read this proposal and tell me three things that would make you hesitate to sign. Be specific.”
Fifteen minutes, start to finish. Proposal quality goes up, not down, because the human spent their time thinking about the prospect instead of wrestling with Microsoft Word.
What This Framework Is Not
A couple of important caveats.
This is not a replacement for domain expertise. The four-part framework helps you get better output. It does not guarantee the output is correct. Everything produced by AI needs a human reviewer — especially anything going to a client, a regulator, or a funder. We’ll cover this in depth in a companion post.
This is not an excuse to paste confidential data into consumer AI tools. The framework applies equally inside an approved enterprise tool and inside a free ChatGPT window. But where you run the prompt matters more than how you write it. If the data tier doesn’t match the tool, better prompting doesn’t save you.
How DVG Systems Helps
Most client teams see a real productivity jump within two weeks of adopting this framework — but only if someone sits down and walks the team through it. That’s what we deliver as part of our Microsoft 365 Copilot rollout service:
- A 90-minute team workshop on the GCSE framework, with examples from your actual workflows
- Role-based prompt libraries — pre-built prompts for common tasks in admin, sales, operations, HR, and finance
- Written quick-reference cards staff can pin at their desks
- Follow-up office hours for the first 30 days, so staff can bring real-world prompts they’re stuck on
- Copilot configuration with the right sensitivity labels and data classification so the productivity gains don’t come at the cost of data leakage
We run this as a packaged service because we’ve found it’s the fastest way to turn a Copilot licence from “another subscription” into real hours saved across the week.
The Bottom Line
You don’t need a prompt-engineering certification. You don’t need to read 50 pages of documentation. You need four slots to fill:
- Goal — what outcome
- Context — why and for whom
- Source — what to draw from
- Expectations — how the output should look
Fill those four, every time, and you’ll be ahead of most AI users — today, with no new tool and no new subscription.
DVG Systems provides managed IT services and Microsoft 365 Copilot rollouts to small and mid-sized businesses across Northern Ontario, including Thunder Bay, Timmins, and the surrounding region. If you’d like a prompt-training workshop for your team, book a free assessment or reach us at (807) 700-0061 or solutions@dvgsystems.com.