How to Automate Agency Proposals with AI and Zapier

How does an automated proposal system work?

Most closers either spend a couple of hours building a proposal from scratch or about 30 minutes adapting a template. This system removes most of that work. After a discovery call, the closer gets a Slack message with a link to a short form. They fill it in, and the automation updates your records and builds a custom proposal using AI.

The whole build has five parts, and the same structure works in whichever tools you prefer. We used Airtable, Zapier, Slack, Google Forms, ChatGPT and Google Slides. You could do the same with Make.com, ClickUp or GoHighLevel.

How do you set up the sales meeting database?

This is a sales meeting hub rather than a full sales hub. A free Airtable account is enough. Create two tables.

Keep it simple to start. You can add more fields later, and a separate leads table is a sensible next step.

• Meetings table: meeting name, notes, a closed checkbox, a qualified checkbox, a proposal required checkbox, the date and time, the duration, the assigned closer (linked to the closers table) and the lead's email.
• Closers table: name, the meetings assigned to them, notes and the closer's email. The email is important because the automation uses it to find the right closer.

How do you track sales calls automatically?

The first automation runs whenever a call is booked. We built it in Zapier with a Calendly trigger for new invitees, but any scheduling tool works, and a round robin setup handles several closers.

Because the trigger fires for every meeting, the first step is a filter so only sales calls continue. We filtered on the meeting name containing the name of the sales call event.

1. Trigger: a new event in your scheduling tool, using invite created.
2. Filter: continue only if the meeting name contains your sales call name.
3. Airtable, find a record: search the closers table by email, using the host's email. Do not use the invitee's email, because that belongs to the prospect.
4. Airtable, create a record: add a meeting with the name, start time, duration, the closer's record ID and the lead's email.
5. Slack, send a message: tell the closer a new sales call is booked, with the lead's name, email and time. A direct message is better if you store each closer's Slack ID in Airtable.

What should the post call feedback form ask?

We built the form in Google Forms, though Tally or Typeform would also work. It does two jobs: it updates the meeting record, and it feeds the AI that writes the proposal.

Questions to include: was the lead qualified (yes or no), did you close the lead (yes or no), do you need to make a proposal (yes or no), what problems the lead had, what goals the lead had, what consequences the lead faces if they do nothing, and the lead's email so it can be matched to the meeting.

For a detailed proposal you will need more questions, so the AI has more to work with. It is also worth making the form conditional, so a closer who closed the lead does not have to answer everything.

How do you make sure closers fill in the form?

Getting closers to update the CRM after every call is a common headache, so this automation sends the reminder for them. We built it inside Airtable, though most tools can do it.

1. Trigger: when a record matches conditions, in the meetings table, where the date is on or before today. Airtable fires this once as each record starts matching, so it triggers as the call time passes.
2. Action: send a Slack message asking the closer to fill out the form, with the form link, the closer's name and the client's name as variables.
3. Optional: store each closer's Slack channel in Airtable and make the channel dynamic, so each message goes to the right person.

How do you generate the proposal with AI?

The last automation triggers when the Google Form receives a new response. It does the heavy lifting.

Before the build, decide whether to write prompts inside the automation or to use a custom assistant. An assistant can be given supporting files, such as your past proposals, so responses are more tailored. For the demonstration we used prompts directly. The prompts are examples, so edit them to fit your offer.

1. Trigger: Google Forms, new form response.
2. Airtable, find a record using the lead's email from the form, then Airtable, update a record to tick qualified and closed and to paste the problems, consequences and goals into the notes. Yes and no answers may need converting to true and false first.
3. Filter: continue only if the proposal answer contains yes.
4. ChatGPT, three conversation steps: one to break down the problem, one for the consequences and one for the goal. Each prompt includes the relevant form answer. Tell it to output only the final text, with no questions or preamble.
5. Google Slides, create presentation from template: name it after the prospect, and map the three AI replies and the client name into the template.
6. Slack, send a message to the closer with the link to the new proposal.

How do Google Slides placeholders work?

In your template deck, write each placeholder inside double curly brackets, with a label such as client name, problem statement, consequence statement and goal statement. The automation then lists each label as a field you can fill with data from earlier steps. The same approach works in Google Docs.

Check the output after a test run. Long AI text may overflow a box, so you may need to shorten the prompt or change the text size in the template. Treat the first results as drafts, and expect to refine the prompts over time.

How can you take the proposal system further?

Send the proposal to the closer first, because AI can make mistakes and you do not want errors reaching a hot lead. Once you trust the prompts, you can send it straight to the prospect by email.

There are three upgrades worth considering. Add a separate leads table to track name, stage, assigned closer, notes, potential value and company. Put the proposal inside a contract with terms the prospect can sign straight away, which removes extra steps from the sales cycle. And if you record sales calls, use a transcript instead of the form: transcribe the recording, have AI pull out the problems, whether the lead qualified and whether a proposal is needed, then run the same automation. The video above walks through the full build.