The Art of the Prompt: Five Elements of Effective AI Communication

If there is one skill that will determine how much value you get from Microsoft Copilot, it is the ability to write effective prompts. A prompt is simply the instruction you give to Copilot — the text you type (or speak) to tell it what you want. The quality of Copilot’s response is directly proportional to the quality of your prompt. Vague, incomplete prompts produce vague, incomplete responses. Specific, well-structured prompts produce specific, high-quality responses that are genuinely useful for legal work.

Writing effective prompts is a skill, and like all skills, it improves with practice and with understanding the underlying principles. This lesson introduces the five-element framework for crafting prompts that consistently deliver high-quality results. Once you internalize this framework, you will find that your prompts become more effective almost immediately, and that the quality of Copilot’s outputs improves dramatically.

The Five Elements of an Effective Prompt

Element 1: Goal
The goal is the most fundamental element of any prompt — it tells Copilot what you want it to produce. Be specific about the type of output you need. “Draft a letter” is a goal, but “Draft a demand letter” is a better goal, and “Draft a demand letter seeking payment of an outstanding invoice” is better still. The more precisely you define your goal, the more accurately Copilot can fulfill it.

Common goal types in legal practice include: summarize, draft, review, compare, rewrite, translate (into plain language), extract, organize, and analyze. Each of these verbs implies a different type of output, and using the right verb helps Copilot understand what you are looking for.

Element 2: Context
Context tells Copilot the background information it needs to generate a relevant response. Without context, Copilot must make assumptions about your situation, and those assumptions may not match your actual needs. With context, Copilot can tailor its response to your specific circumstances.

For legal work, context typically includes: the type of matter (personal injury, contract dispute, estate planning), the jurisdiction, the stage of the proceeding, the relationship between the parties, and any relevant background facts. You do not need to provide a comprehensive case summary — a few sentences of context can dramatically improve the relevance of Copilot’s response.

Element 3: Source
The source element tells Copilot which specific documents, emails, or data to use as the basis for its response. This is what makes Copilot’s responses grounded in your actual matter rather than generic. When you reference a specific document — “using the attached deposition transcript” or “based on the Smith contract in this folder” — Copilot accesses that document through Microsoft Graph and uses its content to inform the response.

Providing sources is particularly important for legal work because it reduces the risk of hallucination. When Copilot is generating content based on a specific document you have provided, it is much less likely to fabricate facts or citations than when it is generating content from its training data alone. Always provide relevant source documents when you need factually grounded responses.

Element 4: Expectations
Expectations tell Copilot about the format, length, tone, and style of the output you need. Legal writing has specific conventions — formal language, precise terminology, structured formatting — and you can communicate these expectations explicitly in your prompt. “Format as a bulleted list,” “write in plain language suitable for a non-lawyer client,” “use a professional but firm tone,” and “limit to one page” are all examples of expectation-setting that will improve the relevance of Copilot’s output.

Element 5: Constraints
Constraints tell Copilot what to avoid or what limitations to respect. “Do not include legal citations in the body text,” “do not speculate about the opposing party’s intentions,” “do not include any information that is not in the source documents,” and “avoid using legal jargon” are all examples of constraints that can prevent Copilot from generating content that is inappropriate for your purposes.

Putting It All Together: An Example

Here is an example of a prompt that incorporates all five elements for a common legal task:

“Draft a two-page client memo [GOAL] explaining the statute of limitations for personal injury claims in California [CONTEXT]. Use the attached case file and California Code of Civil Procedure Section 335.1 [SOURCE]. Write in plain language suitable for a non-lawyer client who is anxious about their case [EXPECTATIONS]. Do not include legal citations in the body text, and do not speculate about the outcome of the case [CONSTRAINTS].”

Compare this to the weak prompt version: “Write a memo about the statute of limitations.” The five-element version will produce a draft that is immediately useful; the weak version will produce something generic that requires extensive revision.

Weak vs. Strong Prompts: A Comparison

The following table illustrates the difference between weak and strong prompts for common legal tasks:

The Iterative Approach: Prompting as a Conversation

Even with the five-element framework, your first prompt will not always produce a perfect result. That is completely normal and expected. Copilot is designed to be used conversationally — you can refine, redirect, and build on previous responses to progressively improve the output.

Think of prompting as a conversation with a highly capable but literal assistant. If the first draft is too formal, say “Make this more conversational.” If it is missing a key point, say “Add a section addressing the indemnification clause.” If the format is not quite right, say “Reformat this as a numbered list with sub-bullets.” Each refinement brings the output closer to what you need, and the conversation can continue until you have something you are satisfied with.

The iterative approach is particularly valuable for complex documents where the requirements are difficult to specify completely in a single prompt. Start with a solid first prompt, review the output, identify what needs to change, and refine. Most legal professionals find that two or three iterations produce a draft that is 80-90% ready for final review and editing.