The goal of this project is to fully integrate AI into Hepworth's sms lead follow-up process. By automating all text conversations through the integration of chatGPT with Twilio, we can centralize all conversation data into Airtable and get one step closer to removing Podium entirely. This will allow for efficient and effective management of all text conversations and automated follow-up messaging.
Project Outline
What are the problems we hope to solve with this project?
- Text messaging happens in multiple softwares
- Canât create customized GPT prompts for specific scenarios
- Have to manually send no response follow-ups
What are the Goals to create in order to solve the problems?
- Automate all text conversations by integrating chatGPT with twilio to handle all conversations, fully managed from Airtable, and automate text send outs follow-up. Centralize all text conersation into airtable to remove the need for podium.
Goals & Deliverables
Create functionality for new Incoming Twilio SMS â GPT + Airtable Log
Goal Planning Tasklist
Deliverables
- When a new sms is received from twilio it should call the endpoint. The endpoint should find the corresponding databse item which should have an airtable record id and conversatoin history stored in gpt format. Using the gpt formatted data it will generate a response to the text message. It will then pull the airtable record id to determine if it can automatically respond to the user. IF yes then it will send the text message, if no then it will store the text message in the matching airtable record in the field SMS Message.
Functionality Flow
Tasklist - Research / Outline / Client Question
Tasklist - Final
- Find number in database
- Using record id find client card in airtable
- If record id doesnât exist then search airtable for matching phone number and use that record id
- if not then create record id, need to make sure you have a record id for the next steps,
- Store incoming sms in airtable interactions table using client record id
- Get
GPT Chat Prompt
from client record and parse for client data variables and replace with client information (use airtable record format {Field Name}) - Combine it with clients Conversation History to generate gpt response
- check field
Can GPT Respond
if YES - calls endpoint to app to send sms
- sends twilio message
- adds to interaction table
- updates airtable clients field
Conversation History Raw
- updates airtable clients field
Conversation History
- updates airtable clients field
Last Message Sent Date
- stores conversation history raw and conversation history and number in APP to quickly pull to generate future responses
- store record id in conversation
- check field
Can GPT Respond
if NO - update airtable clients field
SMS Response
- updates airtable clients field
Last Message Received Date
- stores conversation history raw and conversation history and number in APP to quickly pull to generate future responses
- store record id in conversation
Send SMS From Airtable
Goal Planning Tasklist
Deliverables
- In airtable a person can type a message on a clients card i the field
SMS Response
and then click a button on the cardSend SMS
and it will send the message to the phone number and store the conversation in t
Tasklist - Research / Outline / Client Question
Tasklist - Final
- Create
SMS Respons
field in airtable leads card - Create
Send SMS
button in airtable leads card that - calls endpoint to app to send sms
- sends twilio message
- adds to interaction table
- updates airtable clients field
Conversation History Raw
- updates airtable clients field
Conversation History
- updates airtable clients field
Last Message Sent Date
- stores conversation history raw and number in app to quickly pull to generate future responses
- store record id in conversation
No Response Follow-up + Initial Interaction send SMS + GPT
Goal Planning Tasklist
Deliverables
- Need to be able to send automated follow-up messages to clients who need to complete an action or respond but havenât yet as well as send initial messages when a lead is added from a form submission or some other action to airtable.
Tasklist - Research / Outline / Client Question
Tasklist - Final
Functionality flow for automated follow-up:
- In Airtable create view in Clients table:
Needs Follow-up
- Filter by fields:
Has Client Responded
is 1 andnext action done by
is Automated GPT Follow-up SMS - Create airtable automatoin that runs ever 15 minutes:
- pulls records from
View: Needs Follow-up
- sends client records to api endpoint listed above
Functionality flow for initial send out:
- Create airtable automation that runs on next task done by automated sms or Automated GPT Follow-up SMS
- sends client record to api endpoint listed above
Then combined functionality flow:
- Setup webhook to receieve client ecords
- finds matching record id in app database
- If next action done by is âAutomated GPT Follow-up SMSâ then:
- Pull next action prompt and parse for client data variables and replace with client information (use airtable record format {Field Name})
- uses next action prompt and conversation history to generate GPT response
- sends twilio message
- adds to interaction table
- updates airtable clients field
Conversation History Raw
- updates airtable clients field
Conversation History
- updates airtable clients field
Last Message Sent Date
- stores conversation history raw and number in app to quickly pull to generate future responses
- store record id in conversation
- if next action done by Automated sms:
- sends twilio message
- adds to interaction table
- updates airtable clients field
Conversation History Raw
- updates airtable clients field
Conversation History
- updates airtable clients field
Last Message Sent Date
- stores conversation history raw and number in app to quickly pull to generate future responses
- store record id in conversation
Launch test and production
Goal Planning Tasklist
Deliverables
Tasklist - Research / Outline / Client Question
- Need to research the time it would take to launch an initial test into H&A CRM T and then the live version
Tasklist - Final