How to get AI-drafted sales emails that sound like you

The short answer. A drafted email sounds like its sender when the writer has the customer's full history (past meetings, the email thread, Slack) and instructions for how the sender handles that kind of situation, such as onboarding a new customer. In Day AI the history comes from the customer's conversations in the graph, and the sender writes down how they handle the situation, so the draft follows the sender's judgment as well as their phrasing. Email in the graph comes from Google Workspace (Gmail and Google Calendar); Outlook mail does not sync into it.

What the video covers

• 0:46 A good email takes time because the writer has to recall the last meeting, the earlier thread and what the customer's company does, and a model needs the same material to write well.
• 1:38 A draft has to read like its sender before the sender will trust it.
• 2:33 Mike Peachey, who onboards Day AI customers, shows Day AI combining two kinds of context: everything from the customer's meetings, emails and Slack threads, and written instructions that describe a situation or business process.
• 3:40 He writes instructions for onboarding emails covering data import, integrations, using Day AI in Slack and in Claude, and setting up the sales pipeline, then answers the follow-up questions Day AI asks.
• 5:14 Sign-offs and favorite phrases can go in as instructions; what matters more is how the sender behaves in the email, whether pushing the deal forward or working on the customer's timeline.
• 6:30 With the history and the instructions in place, he reviews the draft and adjusts any line before it goes out.

Excerpts

[0:02] Narrator: Have you been using ChatGPT or Gemini or whatever you find lying around the internet to try to write your emails for you? If you're anything like most people I talk to, you're getting a result that's kind of interesting. Maybe it saves you some time, but it's not really it. "Technically, yes, but actually, no." I'm going to give you three reasons why that's the case. […]

[0:40] Narrator: The second reason that your emails kind of suck is that they don't have context. When you sit down to write an email to somebody, it takes a minute. And if you want to do a really good job, it might take you half an hour. Why is that? Because you have to load up into your mind's eye everything that's going on. You might have to pull up a previous meeting. You might have to pull up the previous email thread. You might have to look and remind yourself what the company does again. We've all been there.

[1:15] Narrator: There's only so much information we can hold in our heads at any given time. And that's kind of the problem: we are holding all of that information in our heads. And if we want to write the email, or we want an LLM to write the email, we have to remind them of all of this context for them to have any shot of doing a good job.

[1:38] Narrator: The third thing is tuning the system just for you. So even with the perfect model, even with all of the context magically ready for the LLM to draft the email, you still need some way to sound like you. For us to feel comfortable looking at a draft email and having the terrifying moment of hitting send, we need to be looking at something that reflects us. It doesn't sound like an LLM, and it doesn't sound like another human. It sounds like us.

[2:10] Narrator: Those are the big three. There's one more detail that's super key that any system you use has to have. You'll see why. First, I want you to hear from Mike Peachey. This guy's talking with dozens and dozens of customers a day. Let's check out how he grinds through his workflow in Day AI like he never imagined would be possible.

[2:30] Mike Peachey: So what we're doing with Day AI is we are automatically stitching all of the context together. And that's in two different modes. We're stitching context across your customer interactions: all of the meeting transcripts, emails, Slack threads, any other data that you give our system about your customers or what they're doing. And we take all of that context and mix it with what we're doing here, creating a template, which is essentially giving context about a situation and how you want to handle it, or context about a business process and what's important about it.

[3:05] Mike Peachey: So we take that and we merge these two things together, and that's how you're able to get a draft that is appropriate for the situation and the context, that uses all of the information we know about a customer. So I can do this really fast and not need to go look up a million things or sit here and crunch all the context in my head. I can get an extremely high-quality draft out that sounds exactly like me, very, very fast.

[3:30] Mike Peachey: In this situation, what I want to do is I want to help create an email template that I can use with new customers during their onboarding. Now, I do a lot of onboarding of new customers. And as a part of that, I'm doing video meetings with them where I'm sharing tips and tricks on how to use Day AI. I'm helping them import their data. I'm helping them get all of their integrations set up. I'm showing them how to use Day AI in Slack or how to use it in Claude. I'm helping them create their sales pipeline, giving them a sense of really how to use the product. And I'm making sure that they've covered enough surface area with it to where they're going to be a retained, happy customer.

[4:08] Mike Peachey: So what I want to do is I want to tell Day AI all of the specific things that are really important to ensure the success of a new customer. And so I'm going to give details on each of those things. And as a part of this, this is now teaching Day AI that when we are in this particular situation with a new customer and we're going through onboarding, when we write a message to them, we're going to refer to this template or this instruction set, and this is going to be the important things that we're considering, and this is the ethos of what the message is that we're going to be sending to this person.

[4:47] Mike Peachey: All right. So now I've given a bunch of context to Day AI about this onboarding business process that we do and what's going to really need to be important about this type of communication. So now, again, Day AI confirms its understanding, and it comes back and it asks me a couple more questions. I answer them. This is a part of what actually helps these emails sound like me.

[5:17] Mike Peachey: You know, when you write an email, it's not just about the little phrases that you use or the way you sign off. You can tell the AI about all of those specific things: phrases you like to use, how you like to start your emails, how you sign them off. Call me Michael in this situation and call me Peachey in another. You can give it all of those instructions, but I think the thing that really matters is how you get across ideas in specific situations.

[5:50] Mike Peachey: How do I actually behave in an email? Am I trying to drive things forward? Am I understanding and looking to take things on a customer's timeline? These are all of the things that matter when it comes to making it actually sound like me. More than just sounding like me: behaving like me. Here's what's so cool about this. […] The system has all of the context. It knows everything about these customer interactions that I've had […]. It also knows exactly how I'm trying to behave and drive the conversation forward in this specific instance, and what's important to our business. And it sounds like me.

[6:30] Mike Peachey: I can tweak and make sure that every single thing in here is the way that I want to deliver this message, and that everything is designed around the way that I actually communicate. And now all I need to do is press send.

[6:49] Narrator: You see that last thing he did in there? That final requirement: whatever you're doing this in, you have to be able to send the email right from there. Move on to the next thing. Boom. Boom. Boom. Get through your work. Get back to being you in real life. Thanks for checking it out. You can come get it for yourself at day.ai.