robocorp.com Open in urlscan Pro
2606:4700:10::6816:2841  Public Scan

Form analysis
0 forms found in the DOM

Text Content

/

DocumentationPortalAPIReMark 💬✨Control Room

DocumentationPortalAPIControl Room

Back to All

Getting Started

Enterprise Platforms

Browser Automation

Desktop Automation

Document Automation

Generative AI

Data and ML

Reporting

Best of the Rest
Robot



OPENAI LLM EMAIL ASSISTANT

Summarise and extract structured data from emails using OpenAI gpt-4


Try It Out

View on Github

Robocorp


Share


Learn how to use OpenAI's large language models to summarise, extract
information and suggest replies with emails! We've made this for a specific use
case described below, but with very little changes to the LLM prompts you can
fit this to pretty much any email based context.


USE CASE - HELPING ACCOUNTS RECEIVABLES PAYMENT COLLECTORS

> Imagine working at a large enterprise company selling industrial goods accross
> the globe. Invoices your customers are supposed to pay are typically large, so
> the cash flow impact of a payment date is significant. The team sends out
> massive quantities of messages about approaching due dates and unpaid
> invoices, and need to read replies manually and update the information back to
> accounts receivables solution.

Our automation helps the payment collectors by generating a summary and reply
suggestions as well as a list of invoices that are being discussed, with a
status based on customer's replies. The bot works in email threads: getting it's
inputs as automatically forwarded emails to Robocorp Control Room (that trigger
the bot execution automatically) and then replying back to the same thread. So
essentially nothing needs to be installed on the end user machine to use the
assistant.

This is how it looks in practise:




👉 THIS IS WHAT THE BOT DOES

 * Get an incoming email to trigger a bot (in Robocorp Control Room)
 * Read email contents
 * Call OpenAI gpt-4 for summary, suggested reply and list of invoices along
   with their data
 * Take OpenAI response and create an email body (HTML) out of it
 * Use SendGrid to send the email back to the user's email inbox


PREREQUISITES

In order to run the bot as-is, you'll need the following in place.

 * Robocorp Control Room and Vault connected to your VS Code.
 * Sendgrid account for sending emails. Free accounts were available at the time
   of writing this for limited usage.
 * OpenAI account with access to gpt-4 model.
 * Following Vault entries either exactly spelled as below, or edit the names in
   code to match your own:
   * Vault OpenAI containing entry key that has your OpenAI API key.
   * Vault Sendgrid containing two entries, SENDGRID_API_KEY that has your API
     key and FROM_EMAIL that has the "from" email address.


SETTING THE PROCESS UP IN CONTROL ROOM

When configuring the process in to Control Room, remember to create an email
trigger under Schedule step. Also remember to keep the Trigger Process and Parse
email checkboxes checked.




RUNNING IN VS CODE

The example provides one email as an example. Use that as a template to create
more of your own test work items! When you run the bot in VS Code, remember to
choose the given input work item when prompted.

TECHNICAL INFORMATION

Last updated

July 7, 2023

License

Apache License 2.0

Tags

OpenAI

AI

Control Room

Email

Outlook

Generative AI

LLM


Report a vulnerability

Follow us

Github Twitter Linkedin Youtube

Products

Products overviewControl RoomAutomation StudioDeveloper tools

Resources

BlogCase StudiesWebinarsPublications

Company

About UsPartnersCareersContact Us
©️ 2023 Robocorp
Privacy · Terms