The problem
For a consultant, email is the job. Overnight, client questions, requests and promises pile up across dozens of threads. Mornings go on sorting what matters from newsletters and notifications, writing replies, and trying to remember what was promised to whom.
What the owner asked for was simple: every morning, go through all my email, draft the replies that need one so I can approve and send them, keep a to-do list across all my clients, and learn my clients over time.
What we’re building
- A one-time Outlook connection. It can read his mail and save drafts. It can’t send, delete or move anything.
- A 6am run every morning. It reads what arrived and what he sent since yesterday, sets aside newsletters and notifications, and works out which client each email is about, whether it needs a reply, and whether it’s urgent.
- Drafts in his voice, in Outlook. Replies are saved into his Outlook Drafts folder, in the same conversation, written from what it knows about that client and matched to how he writes. He reads, edits and presses Send himself, on his computer or phone.
- A to-do list for every client. Promises and requests in email, like “I’ll send the proposal by Friday”, become to-dos with due dates and a link back to the email.
- A page for each client in his own CRM, so what it learns about each client builds up in one place.
How it will work day to day
He opens his CRM and sees what came in overnight: which clients need him, urgent ones first, how many replies are already drafted and the new to-dos. He opens Outlook, reads each draft, changes what he wants and sends. Anything he already answered before 6am gets no draft.
Why it never sends
The assistant writes; he decides. Keeping the send button with a person means a wrong guess costs a few seconds of editing, never a wrong email to a client. During a two-week trial he reviews every draft, and what he sends, edits or ignores is used to improve it.
Where it’s at
The design is approved and his CRM is set up. Next come the Outlook connection and the morning run, then the two-week trial. We’ll update this page once it has been running. We won’t publish results until they’re real.
