Why Blake doesn't send your emails

A subtle decision in Compose is that when you hit Send, Blake doesn't.

Your mail client does. We hand the draft off, pre-addressed, pre-filled, pre-everything, and your own Gmail or Apple Mail or Outlook takes it from there. The actual SMTP, the deliverability, the bounce handling, the reputation, the threading: all of that happens where it already happens. Blake just got out of the way.

This is a deliberate choice, and it might be the most opinionated thing Blake does. Worth explaining why.

The path we didn't take

The obvious thing to build, the thing every "AI sales tool" demo at every conference builds, is server-side outbound. The user clicks Send, the tool's servers fire off the email, the rep sees a confirmation. Looks great in a demo. Looks great in a screencast.

Here is what you sign up for the moment you do that:

• Per-workspace SPF and DKIM setup, because otherwise your mail lands in spam.
• A bounce handling pipeline, because someone has to deal with the addresses that no longer exist.
• A deliverability practice, or a vendor who is one, because mail from a CRM domain ends up reputationally adjacent to mail from every other CRM that uses the same vendor, including the cold-outreach abusers.
• A spam-classifier roulette wheel, because Gmail and Outlook treat mail from "tool.example.com" differently from mail from "you@yourcompany.com," and you do not get to argue with that.

The first three are doable. They are also several engineers' worth of recurring work, and they pay back in a feature most users would not have asked for if you had asked them.

The fourth one is the killer. The thing your prospect wants to receive is an email from you. Not an email from some service that is sending on behalf of you. Not an email with a "Sent via Blake" footer. Not an email that quietly looks like it came from a tool. They want the one that looks like the ten other emails they have from you.

The way to make sure your prospect gets that email is to have your mail client send it.

How it actually works

When you click Compose on a rated opp, Blake drafts the message: subject, body, the right contact in TO, the rest of the contacts in CC if there are any. You see the draft inline, you can edit it, you can hit Recompose if you want a new pass, you can Discard.

When you are happy, you click the button to hand off. Your default mail client opens with everything pre-filled, including a hidden BCC to a Blake address with the opp id baked into the local-part. You read it one more time in your own client, the place you read every other email you write, and you send it from there.

Blake never touches your SMTP. Blake never holds your credentials. Blake never tries to look like you.

The BCC is the only trick. Your sent mail loops back through Blake's inbound pipeline, gets tagged outbound, attaches to the originating opportunity, and gets attributed to you in the activity feed. From your perspective: you wrote an email, you sent an email, and it showed up on the deal. From the receiving side: it is an email from you, full stop.

What this costs you

One click that opens a mail client window. That is the price.

What it buys you: every email from Blake-assisted compose is, deliverability-wise, indistinguishable from every other email you send. Your existing reputation does the work. Your existing threads stay threaded. Your existing filters keep working. The day Gmail rolls out the next anti-bulk-mail change, you do not have to wait for Blake to ship a fix.

And the BCC keeps the activity feed honest. Coffee is for closers. The sent email goes on the deal. Blake gets out of the way.