Booked call research with confirmation email and SMS

Booked call research with confirmation email and SMS

An n8n workflow for sales calls that people book. For each booking in a Google Sheet, an AI agent researches the company on
the web and matches your products. A second agent then writes a confirmation email, a subject line and a text message from
your own success stories. Everything is written back to the sheet, and nothing is sent.

What is inside

• workflow.json: the workflow as n8n exports it, thirteen nodes and no sticky notes. It holds no credential, and the
  Tavily key that sat in one header is replaced by a placeholder.
• README.md: these notes.

What we changed from the original copy

• The key in the Authorization header of Tavily is now <YOUR_TAVILY_KEY>.
• The author's sheet and its three tab ids are gone. The five Google Sheets nodes ask for YOUR_SHEET_ID and find their
  tab by name: Meeting Data, Products or Success Stories.
• The writing prompt signed off with its author's name and company, in the template, the example and the text message. They
  are now YOUR_NAME and YOUR_COMPANY.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An OpenAI API key that can use gpt-4.1.
• A Tavily account and its API key.
• A Google account, and a Google Sheet with three tabs (below).

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Make the sheet with the three tabs below, then copy its ID: the long part of its address between /d/ and /edit.
3. Add a Google Sheets OAuth2 credential. In Review Calls, Update Sheet, Update Sheets 2, Product List and
   Testimonials Tool, select it and replace YOUR_SHEET_ID with the ID. Each node finds its tab by name.
4. Add an OpenAI credential and select it in OpenAI Chat Model and OpenAI Chat Model1.
5. In Tavily, replace <YOUR_TAVILY_KEY> in the Authorization header with your key. The brackets go too, so the header
   reads Bearer, a space and the key.
6. In Sales Writing Assistant, replace YOUR_NAME and YOUR_COMPANY in the template and the examples with your own.
7. Add one booked call to Meeting Data, a product or two to Products and a story to Success Stories.
8. Save, then press Execute workflow at the bottom of the canvas.

The sheet

| Tab | Columns |
|---|---|
| Meeting Data | Call Date, Name, Email, Company Name, Website, Business Type, Project, then company_overview, tech_stack, company_updates, primary_solution, solution_2, solution_3, email_subject, email_text, sms |
| Products | Your catalogue, one row for each product, with plain column headers the agent can read |
| Success Stories | One row for each client story. The writing prompt asks for the person's first name, the company, its business type, the result, the outcome and the solution used |

Fill the first seven columns of Meeting Data for each booking and leave the other nine empty: the workflow writes them.
Spell the headers as above, because the two update tools match them by name.

How a run goes

1. When clicking ‘Test workflow’ starts Review Calls, which reads the rows of Meeting Data whose company_overview is
   empty. A row that already has a company overview is skipped, so you can run it again after adding bookings.
2. AI Agent takes one row at a time. It searches with Tavily, looks up your products with Product List, and must end
   by calling Update Sheet with six fields: company_overview, tech_stack, company_updates, primary_solution,
   solution_2 and solution_3. A structured parser holds its reply to those six.
3. Sales Writing Assistant receives the row and those six fields. It looks for a similar client in Testimonials Tool,
   writes a subject, an email and a text message, and saves them as email_subject, email_text and sms through
   Update Sheets 2. A second parser holds its reply to subject, email and text_message.

Both update tools find the row by Email, so keep one row for each address.

Change the look

• Voice and length: the system message in Sales Writing Assistant. It asks for a short, personal message in a plain
  voice, and gives a template, an example and a text message to follow.
• What the research covers: the system message in AI Agent and the JSON body of Tavily. The body asks for news
  results, three at a time, with an answer and the raw page text.
• Your offer: the rows of Products and Success Stories. The agents only know what those tabs and the search hold.
• Model: OpenAI Chat Model and OpenAI Chat Model1 both use gpt-4.1.

Good to know

• Nothing is sent. The email and the text message only land in the sheet, so add your own sending node after
  Sales Writing Assistant when you want that.
• The writing agent's input has no Call Date, although its template mentions the meeting date. To put the date in the
  message, add a line such as Call Date: {{ $('Review Calls').item.json['Call Date'] }} to the text of
  Sales Writing Assistant.
• Its prompt lists an Email Agent as a tool. The file has no such node, so the agent writes the email itself.
• The first agent is told to call Update Sheet and not to answer with JSON alone. If a row stays empty, open the run and
  check whether the tool was called.
• Each row costs two gpt-4.1 agent runs and as many Tavily searches as the agent decides to make. Try two rows first.
• The prospect's name, email, company and project go to OpenAI, and the agent's search words go to Tavily. Check the rules
  for contacting people where you and they are before you send anything.
• The Tavily key sits in a header, so take it out before you share the workflow.