AI follow-up tracking system, no CRM

The promise that dies is never the big one. It is the small one, made at the end of a call, that nobody wrote down anywhere a CRM would ever see it.

A follow-up tracking system has one job: notice when someone, often you, promised to do something, and surface it again before it goes stale. A CRM does this for deals that already have a stage and a pipeline. Most promises are not deals. They are a line at the end of an email or a call: send that doc over, circle back next week, introduce those two people.

This workflow catches those instead of relying on you to remember to open a CRM and add a task nobody else will ever see.

The failure mode is always the same shape: the promise felt small enough at the time that it did not seem worth a task in a formal system, and then three weeks pass and it is gone, not because anyone forgot on purpose, but because nothing ever asked to be remembered.

What triggers it

A promise appears in the material the system already reads: an email thread, a meeting transcript, or a note you file directly. No separate step to log it.

What the system does

1. Reads the thread or transcript for a commitment

Not a keyword search for "follow up." It reads the exchange for the shape of a promise: someone said they would do something, by when, for whom, matching the same read applied in automatic meeting notes and email triage.

2. Logs it against the person and the project

The commitment gets one line, dated, attached to the contact it involves and the project it belongs to, not a standalone task list disconnected from why it matters.

3. Resurfaces it before it goes stale

On the cadence you set, weekly or in a daily pass, anything still open and approaching or past its implied date gets pulled back into view rather than sitting silently in a file you have to remember to open.

4. Drafts the nudge, does not send it

If a follow-up is overdue, the system drafts the check-in message. It stays a draft until you approve it, the same rule that holds across every workflow here.

A concrete version: on a call, you tell someone you will introduce them to a contact who could help with their search. It is a genuine offer, not a formality, but it is also not the point of the call, so it never gets written down anywhere formal. Three weeks later you run into the person again and realize the introduction never happened, not because you changed your mind, but because nothing ever reminded you it was still open. This workflow exists specifically for that gap between a real intention and a system that would have caught it.

What lands in front of you

A short, dated list of open commitments, whose they are, and how overdue each one is, plus a drafted nudge for anything that has gone quiet too long.

What you still decide

Whether a flagged commitment is actually still live or has been superseded, and whether each drafted nudge goes out as written. The system does not chase anyone on your behalf; it makes sure you know who is owed what.

Why this is the most transferable workflow on the site

Every role that runs relationships rather than a single pipeline hits this same problem: a consultant with many clients, a founder with many investor threads, a recruiter with many candidates. See Claude Code for consultants for the fullest version of this, and pair it with the personal CRM alternative for contact memory.

What makes it transferable is that it does not assume a pipeline shape at all. A sales CRM assumes every relationship moves through the same stages toward the same outcome. Most of the promises worth tracking in a normal working life do not fit that model: a favor owed, an introduction promised, a document someone is waiting on. This workflow tracks the promise itself, independent of whatever process, if any, it belongs to, the same inbox-agnostic instinct behind inbox zero with AI.