Copilot: Build an Email-Writing AI

Key insights

• Copilot AI agent: The video shows how to build a custom Copilot AI agent that drafts emails in your personal writing style.
  It focuses on making messages sound like you instead of generic AI output.
• Copilot Chat: Start by using Copilot Chat to analyze representative emails from your Sent folder.
  The chat extracts tone, phrasing, openings, sign-offs, and structural habits.
• Custom Instructions: Convert the analysis into a single block of Custom Instructions that the agent follows when writing emails.
  Include tone, typical openings, paragraph length, recurring phrases, and formatting rules.
• Knowledge-limited agent: Limit the agent’s knowledge source to email content only so it copies your style and avoids generic or irrelevant content.
  This reduces hallucinations and keeps replies contextually accurate.
• Outlook sidebar: Use the agent inside Copilot Chat or directly from the Outlook sidebar to draft and edit messages.
  This integrates the agent into everyday workflows for faster, consistent drafting.
• Voice replication: The main benefits are time savings, consistent tone, and emails that feel authentic.
  Be aware features depend on your Microsoft 365 license, tenant policies, and security rules, so review governance before deployment.

Overview

Scott Brant’s YouTube tutorial walks viewers through creating a personalized email drafting assistant using Copilot in the Microsoft 365 environment. In plain terms, he shows how to teach a Copilot Agent to draft messages that reflect an individual’s real email voice rather than a generic AI tone. The video is practical and demo-driven, and it emphasizes using real Sent Items to capture typical phrasing, openings, and sign-offs. Consequently, the result aims to read as if the user wrote the email themselves.

Moreover, Brant contrasts this agent-based approach with the built-in drafting features in Outlook, highlighting when an agent brings added value. He points out that agents can be configured to use only an email archive as their knowledge source, which helps maintain stylistic consistency. Additionally, the tutorial covers where to invoke the agent inside Copilot Chat and how to draft directly from the sidebar in Outlook. As a result, viewers get both the conceptual background and hands-on steps.

How It Works

The core technique Brant demonstrates begins with a targeted analysis of representative emails. First, you ask Copilot Chat to review samples and produce a unified block of Custom Instructions that describe tone, structure, recurring phrases, and formatting preferences. Then you create a dedicated Copilot Agent that uses those instructions and limits its knowledge to the email content you specified. This pipeline turns descriptive patterns into reusable drafting rules.

Next, you deploy the agent in scenarios where consistent voice matters, such as routine updates or client communications. Brant shows how the agent integrates into workflows by being callable inside chat and accessible from the Outlook sidebar. He also explains that the agent differs from one-off draft generation because it follows the stored instructions every time. Consequently, teams and individuals who send many similar messages benefit from the repeatability.

Finally, the tutorial addresses necessary permissions and tenant considerations, noting that availability may vary by Microsoft 365 license and organizational settings. Brant reminds viewers to check governance and security policies before creating or sharing agents. He also highlights that the agent’s outputs depend on the quality and representativeness of the email samples it analyzes. Therefore, preparation matters: a poor sample leads to weak instructions.

Benefits

First, the video argues that tailored agents increase authenticity by replicating specific wording and phrases that a sender uses often. This matters because consistent tone supports professional credibility and reduces the risk of messages sounding canned or impersonal. Moreover, using an agent saves time on repeatable communication tasks, allowing users to generate drafts quickly and then polish them if needed. As a result, productivity rises without sacrificing voice.

Second, Brant highlights how limiting an agent’s knowledge to email content can reduce irrelevant or hallucinatory output. By narrowing the knowledge source, the agent focuses on actual writing habits rather than drawing on broad internet data. This improves relevance and makes the drafts feel more like the sender’s own language. Additionally, agents can capture structural habits—like when to use bullets versus prose—which boosts clarity and reduces editing time.

Third, the approach scales for teams that need uniform communication standards across many senders. When organizations want to preserve a brand or internal style, agents provide a repeatable rule set that different people can use. Consequently, managers can achieve consistency while allowing individual variations where appropriate. However, scaling requires thoughtful governance and oversight.

Trade-offs and Challenges

Despite the promise, Brant’s method involves trade-offs that editors should weigh. For instance, capturing an authentic voice requires a representative and non-sensitive sample of Sent Items, which raises privacy and compliance concerns. Organizations must balance personalization with data protection and ensure that samples do not expose confidential information. Therefore, governance and clear review processes are essential before deployment.

Another challenge lies in handling diverse audiences. A single agent tuned to a user’s default voice may not suit every recipient, such as executives, external clients, or legal teams. Brant suggests adjusting prompts or creating multiple agents, but that raises maintenance overhead. Consequently, teams must decide whether to favor one consistent voice or to manage several targeted agents.

Finally, the technique depends on platform availability and ongoing maintenance. Features and behavior can vary by tenant, license, and regional rollout, and the custom instructions may need revision as writing habits evolve. Thus, while the initial build can deliver immediate gains, keeping agents accurate requires periodic review and human oversight. In short, the benefits come with work to sustain them.

Practical Steps and Considerations

For editors and IT leads planning implementation, Brant’s tutorial offers a clear sequence: gather representative Sent Items, prompt Copilot Chat for consolidated Custom Instructions, create the agent, and then test it in real drafting scenarios. He recommends starting with low-risk, high-volume communications to validate the agent before wider use. In addition, perform reviews that involve both the sender and a compliance owner to confirm tone and safety.

Operationally, document how agents are created and maintained, and set rules about what data sources they may access. Training for end users also matters because people need to know when to trust the agent and when to edit. Finally, treat the agent as a tool that speeds drafting but does not replace critical review for sensitive or high-stakes messages. This balanced approach reduces risk while preserving the upside.

Conclusion

Scott Brant’s video provides a pragmatic path to making email AI feel more human by turning observed writing habits into reusable agent rules. It blends hands-on steps with cautions about governance and sample quality, so readers walk away with both a how-to and a lens for prudent implementation. For newsroom editors, the takeaway is clear: personalized agents can boost consistency and productivity, but only when paired with careful data handling and ongoing oversight.

Therefore, teams should pilot the method, document outcomes, and refine their approach over time. Meanwhile, maintain human review for sensitive messages and update custom instructions as styles change. Doing so will help organizations realize the benefits that Brant demonstrates while managing the practical trade-offs the tutorial highlights.

Keywords

Copilot AI email agent, build Copilot agent, AI email writer, train AI to write emails, Microsoft Copilot email tutorial, personalized email AI assistant, automate email drafting with AI, create GPT email assistant