Communicate·Advanced·60 min·Updated Sep 30, 2026

Build a personalized customer communications agent

Stand up a Copilot Studio agent that reads a customer's communication history and drafts messaging tailored to how they actually respond.

Download PDF

Microsoft 365

Works With

Prerequisites

Access to customer communication history in your CRM or mailbox

Business Outcome

Customer-facing messages match each account's tone and history automatically, lifting response rates without a rep hand-tailoring every email.

Handle with care

Customer communication history

This agent has standing access to a customer's full communication history to personalize messaging. Scope its data source to what's actually needed for drafting tone-matched messages, not an unrestricted view of every customer interaction.

Workflow Overview

Outlook
Copilot Studio

Step 1: Define the agent's scope and data sourcesCopilot Studio

In Copilot Studio's Knowledge settings, connect the agent to the account's email history (via the Outlook connector) and CRM notes, and explicitly list which message types it's allowed to draft.

Renewal outreach, check-ins, follow-ups — rather than leaving it open to draft anything, which risks it generating something outside what a rep actually asked for.

Step 2: Configure the personalization instructionsCopilot Studio

In the agent's instructions, spell out exactly what "tone matching" means in practice — what counts as formal vs. casual, how far back to look in the history, what triggers the "hasn't responded" flag.

Rather than a single-line instruction like "personalize the message," which leaves the agent to invent its own definition of personalization.

Prompt idea:

When drafting outreach to a customer, review their last [how many] interactions for tone (formal vs casual), typical response time, and topics they've raised repeatedly. Match the draft's tone to their pattern and flag if they haven't responded to anything in [how many] days.

Step 3: Test the agent on a known accountCopilot Studio

Run the agent against an account whose communication style you already know well, and check that the tone and content it produces actually matches what you'd expect for that customer.

Step 4: Route every draft through a rep before sendingCopilot Studio

Configure the agent to produce drafts for review, not auto-send — a rep should confirm each message reflects the account's real current situation before it goes out.

Step 5: Roll out to the full account list graduallyCopilot Studio

Start with a small group of accounts, review a sample of drafts each week, and expand once the tone-matching and fact accuracy hold up consistently.

Check the work

  • Compare a sample of drafted messages against what a rep who knows the account would actually write.
  • Confirm no draft states something about deal status or open issues that isn't current.
  • Check that low-engagement accounts are flagged for a human look rather than getting an automated message anyway.

Source: Microsoft Copilot Scenario Library — Sales (2026)

Expected Outcome

A working Copilot Studio agent that drafts customer messages pre-tailored to each account's communication pattern, for a rep to review before sending.

✓

AI is the right call here

Reading five past interactions to infer whether an account prefers formal or casual tone, and matching a new draft to that pattern, is a judgment call about a person's communication style — not something a mail-merge field can hold.

Related Workflows

Related Playgrounds

WORK WITH LIMINALS

Ready to roll this out beyond one person?

Workflows like this tend to raise real governance and licensing questions once more than one person is using them — that's exactly what we help with.