Copilot: 10 Prompting Rules That Work

Key insights

• Set the right tone — Treat Copilot like a capable colleague, not a servant.
  Start prompts with a clear goal and use the formula Action verb + quantity + audience (for example: “Create a 5‑bullet executive summary for the board”).
• Context and Specify Sources — Give timeframe, role, and files so Copilot picks the right data.
  Point to exact documents or say “use only the attached files” to reduce errors and make outputs auditable.
• Constraints and Structured output — Define length, tone, format, and acceptance criteria up front.
  Ask for tables, JSON, or bullet checklists when you need machine‑readable or ready‑to‑use results.
• Draft first and Power phrases — Ask for a quick draft, then refine; explain the "why" behind requests to improve relevance.
  Use power phrases like “think step by step” or “review and improve” to activate deeper reasoning.
• Scaffold and Brevity modes — Provide templates or examples to guide output and choose how concise you want it (expert, brief, simplifier).
  Scaffolds cut back-and-forth and speed repeatable results.
• Modular prompts, Copilot Studio, and Security Copilot — Break big tasks into reusable steps, use Copilot Studio features (prompt builder, testing, Power Fx) for repeatability, and prefer “You” instructions in security or action prompts.
  Modular prompts improve accuracy, reuse, and audit trails for enterprise workflows.

In a concise masterclass-style YouTube video titled "The 10 Copilot Prompting Rules That Actually Work," Daniel Anderson [MVP] walks business users through practical prompt engineering strategies for Microsoft Copilot. The video aims to help professionals avoid what he calls AI slop by showing how small changes in phrasing and structure produce more reliable outputs. Throughout the presentation, he mixes concrete examples with a full worked prompt, and he highlights how to apply these techniques across Microsoft environments. As a result, the guidance is both tactical and suited for enterprise scenarios where accuracy and repeatability matter.

Overview of the Rules and Structure

Anderson organizes the lesson around ten clear rules that span tone, scope, formatting, and task design, and he timestamps each section to make it easy to follow. He opens by urging viewers to treat Copilot like a capable colleague rather than an obedient tool, which sets the tone for collaborative prompts that ask for work to be done rather than issued as commands. Then he proceeds to practical rules such as using explicit action verbs, specifying quantity and audience, and setting explicit constraints on length and format to reduce ambiguity. By the end, he demonstrates how to combine the rules into a single prompt suitable for producing board-level summaries.

Moreover, the video groups several rules that are especially useful in business contexts, such as naming sources, asking for structured output, and explaining the rationale behind a request so the model reasons in a helpful way. Anderson also highlights techniques like scaffolding prompts with templates, requesting step-by-step thinking, and using personas or power phrases to guide style and depth. These tactics are presented as repeatable building blocks that can be reused across documents, reports, and automated workflows. Therefore, the guidance aims to move users from ad hoc queries to disciplined prompt design.

Practical Application and Tradeoffs

Applying the rules involves tradeoffs between speed and precision: short, informal prompts are fast but often deliver inconsistent results, while explicit, constraint-filled prompts take more time to craft yet yield higher precision. Anderson argues that investing effort up front—by defining the output type, audience, and acceptance criteria—reduces downstream revision cycles, which often offsets the initial time cost. However, teams must weigh the overhead of building templates and modular prompts against the volume of work they ask Copilot to handle, since smaller teams or one-off tasks may not justify a heavy template regime. In contrast, enterprise workflows and security-sensitive tasks typically benefit from the added structure.

Another tradeoff involves creativity versus control: strict constraints like rigid word or format limits improve consistency but can stifle exploratory or generative uses of the model. Anderson suggests a mixed approach where users first request a draft with loose constraints and then refine with tighter acceptance criteria, which helps balance innovation and reliability. This staged approach also supports testing and iteration, so teams can measure which prompt components deliver the most value. Consequently, the method encourages reuse of proven prompt modules while allowing room for experimentation.

Challenges and Limitations

The video acknowledges persistent challenges such as hallucination, incomplete context, and data privacy concerns when Copilot accesses organizational content, and it recommends explicit source directives to mitigate these issues. While telling the system which files to use and which to ignore reduces errors, it also introduces governance questions about access and auditability that IT and security teams must manage. Additionally, product changes and model updates may alter behavior over time, so prompts that work today may need revision tomorrow, creating a maintenance burden for large prompt libraries. Thus, teams should track performance and update templates as Microsoft features evolve.

Further limitations arise when complex tasks require domain expertise or precise logical chains that a single prompt cannot deliver reliably, which is why Anderson emphasizes breaking work into modular subtasks. Dividing and conquering complex assignments allows for targeted verification at each step, which improves overall trust in the final output. Still, this modular approach increases orchestration complexity and may demand workflow automation or simple tooling to manage prompt sequences efficiently. Therefore, organizations should balance modular rigor with practical orchestration capabilities.

Tools, Templates, and Best Practices

Throughout the video, Anderson points to practical features such as using templates, scaffolds, and prompt testing tools within Copilot Studio to create auditable, repeatable outputs that integrate with Microsoft 365. He encourages users to bind prompts to specific data sources, to request structured formats like JSON or tables when downstream automation is planned, and to include clear acceptance criteria so outputs are actionable. Moreover, power phrases like "think step by step" and "review and improve" help coax stronger reasoning and iterative refinement from the model. These tips provide a foundation for operationalizing prompt design in teams that need consistent business outputs.

In closing, the masterclass offers a pragmatic roadmap that balances clarity, control, and reuse, and it stresses the importance of testing and governance as part of any Copilot rollout. For organizations that invest in reliable prompt libraries and simple testing practice, the payoff can be fewer errors and faster production of business documents. Ultimately, Anderson’s video serves as a compact, usable guide for professionals who want to get consistent value from Microsoft's AI assistants while managing the tradeoffs and operational challenges that come with enterprise AI adoption.

Keywords

Copilot prompt rules, Copilot prompting tips, Copilot prompt examples, Microsoft Copilot prompts, Copilot prompt engineering, best Copilot prompts 2025, how to prompt Copilot, effective Copilot prompts