AI email agent: How to draft emails
Davis Christenhuis
Email still consumes hours of every workday. Teams spend that time reading threads, researching context across different tools, and drafting responses that need to be both accurate and on-brand.
AI email agents are built to handle the repetitive parts of that workflow. This guide covers what they actually do, how they differ from basic email assistants, and how to set them up effectively.
📌 TL;DR
Here are the key takeaways:
- What it is: AI email agents connect to your CRM, knowledge base, and past conversations to draft complete emails based on real company data.
- What it does: The agent reads the email messages you share with it, pulls context from your connected tools, matches your writing style, and generates drafts for you to review before sending.
- How Dust works: Build a custom agent, connect it to your data sources (CRM, docs, email), and set clear instructions. Simple agents can be set up without engineering resources. More advanced workflows may involve deeper integrations.
- Why accuracy matters: The best agents use information from your connected systems and point out missing details instead of making things up. Using real data helps reduce mistakes.
What is an AI email agent?
An AI email agent is a system that drafts, responds to, and manages emails by accessing your company's knowledge base, CRM data, and past conversations. It reads the full context of an email thread, pulls relevant information from your connected tools, and generates complete responses based on that information.
In practice, this means the agent can reference a customer's purchase history when drafting a support reply, pull product specifications from your documentation when answering a prospect’s question, or check past meeting notes before writing a follow-up email.
The agent works with the information from your connected tools, pulling together context that would take a human some time to gather manually.
Not every tool actually works this way. Some focus on inbox organization or text prediction without connecting to your systems. Agents built for business workflows pull from real data sources to draft responses grounded in customer history and internal information.
💡 Curious to see what Dust can do for your team? Try it free for 14 days →
What do AI email agents do?
Here's what they're built to handle:
- Analyze email threads: Before drafting anything, the agent reads the email messages you share with it to understand what's been discussed, what questions remain, and what the person on the other end is asking for.
- Connect information across your tools: When a customer asks about their order status, the agent pulls from multiple sources instead of guessing or using a template. It queries your CRM for their account history, checks your support platform for related tickets, and references internal documentation to see if there's relevant context.
- Match your tone and format: The agent references your past emails to pick up on patterns in how you write, including sentence length, word choice, how you open and close messages, and whether you use bullet points or paragraphs. The more examples it can draw from, the more closely its drafts reflect how you actually communicate.
- Generate drafts for your review: The agent creates a full response with all the necessary information formatted and ready to send. You read it, make any changes you want, and decide whether it goes out. The drafting work is done, but you keep final control over the message.
How Dust's AI email agent works
Dust is a platform where you build custom AI agents for your specific workflows. You create agents that connect to your CRM, documentation, and other data sources. These agents can draft emails using that context, and they can also help with many other tasks.
Draft emails grounded in your company knowledge
Most email tools only see what's in the thread. Dust agents understand the full context. So instead of generic drafts, you get emails that reflect your tone, the specifics of the situation, and the person you're writing to, without filling in the blanks every time.
Step 1
Create a new agent in Dust and write clear instructions for what it should do. In this example, the agent is set up to draft professional email replies by reading thread context, referencing connected knowledge sources, and matching your tone.
Connect it to the data sources it needs. This one connects to Notion for company knowledge and Gmail to read conversations and save drafts.
You also choose which AI model powers the agent. Dust supports models from OpenAI (GPT-5), Anthropic (Claude Opus, Sonnet, and Haiku), Google (Gemini), and Mistral, so you can pick the right balance of speed, cost, and capability for your workflow.
Step 2
Send the agent a message with the email you want to respond to. The agent reads the thread, queries your connected data sources for relevant context, and generates a complete draft.
Before saving the draft to your Gmail, it asks for permission. This keeps you in control of what the agent can access and modify.
Step 3
The drafted email appears directly in your Gmail inbox as a reply to the original thread. You can review it, make edits if needed, and send it when you're ready. The agent handles the research and writing. You handle the final decision.
The agent is designed to reference information from your connected sources. If it can't find a confident answer in your systems, it flags the gap rather than guessing. Because responses are based on your real data, the risk of inaccurate or fabricated details is reduced.
This basic example can be set up in minutes. More complex workflows, like the Brevo case below, take longer to configure but the core process stays the same.
💡 See how Dust works for your team. Start your free Dust trial →
How Brevo cut email personalization time by 80% with Dust
Brevo, a Paris-based customer engagement platform serving over 600,000 customers globally, faced a challenge common to scaling go-to-market teams. Sales reps needed to send hyper-personalized emails to hundreds of prospects weekly, but manually researching each one took 30+ minutes per person.
At that pace, personalization meant sacrificing volume. The cost wasn't just time; it was missed revenue moments and inconsistent execution across the sales team.
The company built an email agent connected to their CRM data through Supabase (an open-source PostgreSQL database that mirrors their CRM records), with integrations for LinkedIn and web research tools.
When a sales rep selects contacts to reach out to, the agent enables:
- Automated prospect research at scale: The agent pulls complete CRM history, LinkedIn profiles, company news, and firmographic data for every prospect without manual lookup
- Context-aware personalization: Specialized sub-agents route prospects based on type (gated content downloads get one playbook, cold e-commerce leads get another) and generate three email variants tailored to role, seniority, and context
- Consistent quality across the team: Every sales rep now has access to the same depth of personalization, eliminating the tribal knowledge gaps that used to cause inconsistent outreach across the team
The results: Time spent on email research and drafting dropped by 80-90%, from 30+ minutes per prospect down to a few minutes of review and sending.
💡 Want to see how other teams use Dust? Read more customer stories →
Frequently asked questions (FAQs)
What's the difference between an AI email agent and an email assistant?
An email assistant typically helps you write faster with suggestions and autocomplete based on the text in front of you. An AI email agent goes further: it connects to your CRM, knowledge base, and past emails to autonomously gather context, reason over multiple sources, and draft complete responses grounded in your company's real data.
What should I look for in an AI email agent?
The most important capability is access to your company's actual data. The agent should connect to your CRM, documentation, and past conversations, not just work with the text in the email thread. It should only reference information it can verify, not fabricate details. And it should let you review drafts before sending, so you stay in control of what goes out.
Does Dust work with Gmail and Outlook?
Yes. Dust connects to both Gmail and Outlook. With Gmail, agents can read messages and create drafts. With Outlook, agents can read messages and create drafts. Both integrations let the agent draft replies that appear in your inbox for review. You can also use Dust agents in Slack and copy the drafted email wherever you need it.
Can I customize how Dust agents write?
Yes. You can set explicit guidelines like 'Keep emails under 150 words' or 'Always use a professional tone.' When the agent has access to your past emails, it references them to match your writing style, tone, and structure. The more examples it can draw from, the better it replicates how you actually communicate.