← Field notes// field note

From Slack intake to AI-drafted proposal: automate lead follow-up

Leads go cold when follow-up depends on someone remembering. A Slack pipeline that remembers for you, up to an AI-drafted proposal.

October 2026

6 min read

You can run inbound lead follow-up inside Slack. A form posts the lead, the team claims it by replying, timed rules send the follow-up email, a booked call creates a private deal channel, the meeting summary lands there, and AI drafts the proposal in a Google Doc. A person edits and sends it. Only two emails reach a client automatically.

From Slack intake to AI-drafted proposal: automate lead follow-up, QuikCode field note

Can you automate a sales pipeline in Slack, all the way to a proposal? Yes, and the team barely has to change how it works. A form posts the lead into Slack, teammates claim it by replying "interested", timed rules send the follow-up, a booked call opens a private deal channel, and AI drafts the proposal from everything that was said and sent. A person reads it, edits it and sends it. That last part matters most.

I built this for a US investment and advisory firm that takes in founders looking for fundraising help. Before, every intake landed in a Slack channel and then depended on someone remembering to claim it, chase it, email the founder and follow up after the call. Contact data was spread over a form tool, Airtable, Pipedrive, a newsletter tool, an SMS tool and some old Make scenarios. Nothing was broken, exactly. Things just slipped.

How the pipeline runs, step by step

Everything runs in one self-hosted n8n workflow (209 nodes, for the curious). It does this:

  1. The intake. A founder fills in the intake form. It posts into a Slack channel.
  2. Claim by reply. Every 10 minutes the workflow scans that channel. A teammate claims a lead by replying interested in the thread. No new tool, no new habit.
  3. 48 hours. If nobody has claimed it, the bot posts one reminder to the channel. One, not ten.
  4. 72 hours (or 48 if claimed). The founder gets a personal follow-up email from the managing partner, in one of three versions depending on what they asked for. Interested teammates are copied. The same run posts a summary for the team and adds a note in Pipedrive.
  5. A booked call. When the founder books through Calendly, a private Slack channel for that deal is created, the right people are invited, the intake is posted there and the calendar invite gets the team on it. Pipedrive gets the person and a deal.
  6. No booking after 7 days. A short nudge email goes to the founder, and the team lead gets tagged.
  7. After the call. The meeting summary and recording link land in the deal channel. A day later, teammates who haven't commented get a nudge.
  8. The proposal. A day after the summary, AI drafts a proposal as a Google Doc and posts it in a proposals channel, tagging the partner.
  9. Cleanup. A deal channel with no activity for 30 days gets archived automatically.

Only two automated emails ever reach a founder: the follow-up and the 7-day nudge. Everything else stays internal.

Where the AI sits, and where it doesn't

The AI does one job: write a first draft of the proposal. It reads the intake and the files that came with it (like a pitch deck), the call summary or transcript, the discussion in the deal channel and the email thread with the founder. It writes in the firm's own style, from a long written brief of how they position themselves.

Then a human takes over:

For founders who go quiet there is no AI at all. A plain nudge email does that job.

What happens to the data

Only the proposal step uses AI. It runs through the OpenAI API, which gets the intake, the attached files, the call summary, the deal channel discussion and the email thread for that one deal. Data sent through the OpenAI API is not used to train OpenAI's models.

What it took to make it reliable

The flow is the easy part to draw. Keeping it running is where the time went.

What you can connect

The pattern fits any business with inbound leads, applications or requests that need a human reply:

A similar pipeline idea, for a different business, is in the Deal Engine case. More builds on my work page.

When this is not the right approach

The goal is that nobody has to remember. If leads in your business depend on someone's memory, let's talk for 30 minutes.

Can you build a sales pipeline in Slack?

Yes. A form posts each lead into a channel, teammates claim it by replying in the thread, and a workflow (for example in n8n) handles reminders, follow-up emails, CRM notes and a private channel per deal once a call is booked.

Can AI write client proposals?

AI can write a solid first draft if it gets all the context: the intake, attachments, call summary, internal discussion and email thread. Let a person revise and send it. In this setup the proposal is never sent automatically.

Is the data used to train the AI?

No. The proposal step runs through the OpenAI API, and data sent through the API is not used to train OpenAI's models. It is the only step in the workflow that uses AI.

How do I create a Slack channel automatically when someone books a Calendly call?

Use the Calendly webhook for new bookings. A workflow creates or reuses a private channel for that lead, invites the right teammates, posts the booking details and intake, and adds the team to the calendar invite.

How do I stop leads going cold after a call?

Set timed rules: a reminder when nobody has claimed a lead, a follow-up email after a fixed time, a nudge when no call is booked within a week, and a nudge to teammates who haven't weighed in after the call.

Is self-hosted n8n reliable enough for this?

It can be, with care. Mirror the workflow's state somewhere durable, like Airtable, so a restart can't wipe it, add a monitor per step, and send every error to one channel someone actually reads.

Got a process that should be software?

Book a call
© 2026 QuikCode · software for better operations