OpenAI GPT-3.5 Turbo x Email x make x NocoDB: Low Code Smart Customer Service Helpdesk

Zendesk indeed is a great choice when come to customer service helpdesk, which we covered it here. However is our favorited low code platform NocoDB able to do the same? Most certainly, although it might not provide a comprehensive helpdesk functions like Zendesk, but being an open source options, and small team, it is still able to fit the job well, especially with the help of OpenAI, we can now even built the similar business process flow as what we did in Zendesk’s article.

Ok, so here we are today, we going to built a helpdesk system with NocoDB, with low code of course, and with below requirements

  • Submit a ticket via form
  • Submit a ticket via email – if being identified by OpenAI as a complain

Quick run-through

Pre-requisite check

  • Initial requirement

    Let’s look at how it started, as well as the initial design

  • Backend

    • Email: To receive incoming email for customer service, and to send out the response
    • OpenAI GPT-3.5 Turbo: Served as AI brain which handling the response
    • NocoDB: Served as helpdesk platform
    • Make: Served as integrator which “glue” different system together
  • Platform

    • Email: Be it common IMAP account, or Microsoft Office 365
    • OpenAI GPT-3: SAAS Model
    • NocoDB: On-Premise
    • Make: SAAS Model
  • Account

    • Email: Required any email account that support IMAP / SMTP
    • OpenAI GPT-3.5 Turbo: Required a paid account
    • NocoDB: Just download the docker and get it run
    • Make: Required a basic free account
  • Cost

    • Email: Do consult your IT department, if you are not sure where your email is, or if it is support IMAP / SMTP
    • OpenAI GPT-3.5 Turbo: Required a paid account, you may check the pricing info here.
    • NocoDB: Open-Source.
    • Make: Required a basic free account, you may check the pricing info here.
  • Programming

    That’s a minimal HTML and JSON knowledge required

Applications Flow

  • Email

    Email is the entry point of the customer service

  • null

    OpenAI GPT-3.5 Turbo

    Identify the tone of the feedback, and suggest a response based on the feedback

  • null

    Make

    It is platform that connect different applications, without coding! Other than connect different application using connector, and also manage different routes based on different criteria

  • null

    NocoDB

    Served as helpdesk platform, to handle complain

At the time the article was published, OpenAI just release GPT-3.5 Turbo, hence we are using this as 1 of the tool for the AI part

Table Structure

Header

Name
Type
Mandatory
Length
Remarks
Title
Text
Yes
255
Name
Text
Yes
255
Email
Text
Yes
255
Add New
Content
Long Text
No
Unlimited
Status
Text
No
10
Open / Close
Summary
Long Text
No
Unlimited
For AI usage
Replies
Link Records
Yes
N / A
Link to Replies table, 1 to many

 

Replies

Name
Type
Mandatory
Length
Remarks
Reply
Text
Yes
255
ResponsedBy
Text
Yes
255
Header
Linked Record
N / A
N / A
Linked to header records
SenderNameLookup
Lookup
N / A
N / A
Get from header
SenderEmailLookup
Lookup
N / A
N / A
Get from header
TitleLookup
Lookup
N / A
N / A
Get from header
SenderName
Formula
N / A
N / A
Get from lookup
SenderEmail
Formula
N / A
N / A
Get from lookup
Title
Formula
N / A
N / A
Get from lookup

Let's get started

NocoDB

1. We covered the tutorial on creating a contact form using NocoDB before, you may refresh the memory here. For the helpdesk, let’s start with something new.

 

2. Create the table for Header, let’s start with only below columns

  • Title
  • Name
  • Email
  • Content
  • Status (Open / Close)
  • Summary (going to fill up by AI)

 

3. Next, create only 2 columns for Replies table for now

  • Reply
  • ResponsedBy – At the moment, NocoDB still don’t have any column type for “User”, we have to use text to record user name

 

4. Go back to Header table, create the column “Replies”, this time round the column type set to “LinkToAnotherRecord”, which is link to Child Table, and select “Has Many”

It mean, 1 record in Header table, will be having multiple replies in Replies table.

 

5. Now, let’s head back to Replies table, and create the rest of the column. Now you can see a column called “Header”, which is a link from Header table.

 

6. Let’s move on to create another columns for Name, Email, and Title. These are the look up column from Header table.

We need this column when creating a trigger to send out email upon any reply from our support personnel.

 

7. When we create an email trigger, we actually make use of REST API to trigger. At this point of time when this article is published, NocoDB not able to refer to Lookup columns in REST API trigger, hence we need to create another 3 columns, which is Formula.

The 3 columns are SenderName, SenderEmail, and Title, which simply refer to the lookup columns we created in step 6 just now.

The formula you may written as below sample

{COLUMN_NAME}

 

8. Now, create a form to test it out – just like what we did previously. From the form, you can “hide” the unnecessary fields. Please don’t “delete” it, it will delete the column from the table as well.

 

 

9. You may test it out by submit a sample complain.

 

10. You shall see the entry already logged into Header table. Click on the left side of the record, you shall see a modal pop up slide out from the right.

 

11. Under Replies section, click the button “Link to Replies”, you shall see the replies pop up

 

12. Click on Add new record, you shall see another modal pop up slide from right again, which is the reply form. Please note the slide form are auto generated, which can’t be customized at the moment.

13. You shall have a pretty good idea how to built a simple helpdesk with NocoDB, we will move to make.com, to create a webhook for email trigger. At this point of time, NocoDB not allow to send email to any targetted email yet.

Webhook Email Trigger

1. NocoDB do have SMTP connector, but it is only able to send email at pre-defined event, and pre-defined email. To make NocoDB send email to the sender, we have to use 3rd party solution. Usually you can look at SMTP provider such as AWS SMS, SendGrid, SMTP2GO… those are pretty good SMTP service providers. However, we will do something different, which is using a webhook service in make.com

2. First, create a webhook in make.com under a new scenario. You shall have the endpoint URL ready for the first time, and waiting to listen to incoming REST API trigger to determine the data structure.

Copy down the endpoint URL.

 

3. Head back to NocoDB, under Replies table, add a Webhooks

 

4. Setup the webhook as below, and paste the endpoint URL beside the POST.

You use POST method when you want to send some data over to an endpoint. For the triggering event, we use “After insert”, meaning every reply, we will sent an email to the sender.

The data to sent to the endpoint is easy, NocoDB already prepared for you with a sample, you can edit the way you want. You may make use of my sample below. You can do a test anytime, but using “Test Webhook” to send to the endpoint, or simply add a reply.

 {
    "Reply": "{{data.Reply}}",
    "ResponsedBy": "{{data.ResponsedBy}}",
    "SenderName": "{{data.SenderName}}",
    "SenderEmail": "{{data.SenderEmail}}",
    "Title": "{{data.Title}}"
  }

5. Now, let’s head back to the make.com, you shall see the data structure already been determined, and you can create another connector, which is SMTP.

You will first need to setup the SMTP connectivity with your mail server – you may get IT department help on this.

For the rest, such as To-Email addresses, Title, and Body, you may follow pretty close as below.

6. Alright! You just got a fully function “simplified” helpdesk which able to send an email response to sender.

OpenAI GPT-3.5 Turbo

1. We now having a “simplified” helpdesk system, with email capability, now is the time to add some “intelligent” to it by using OpenAI GPT-3.5 Turbo. You may refer to the business flow here.

We already created the scenario with Zendesk before, you may take a look here. We shall clone it to save our time.

 

2. Once we done the cloned, we shall replace the Zendesk with NocoDB connector, and refer to the Header table we created earlier ago. The configuration you may follow as below.

What? You not sure how to setup the NocoDB connectivity? Fear not, take a look here.

 

3. We shall done for the entire setup, let’s give it a try.

Let's try it out

1. The entire setup in Make should look like below

 

2. You may use the sample complain and compliment email content from here.

3. In any case, for Complain case, here is the NocoDB screen should look like

 

4. Here is the catch, when you click on the Link to Replies, and think to Add new record for the reply, and you can see the rest of Replies records for you to link. This is not a bug, this is the nature of how the link records should work.

As a helpdesk support personnel, you shall always Add new record to create new reply.

 

5. Under Replies table, you shall see all the replies being linked with correct header, and the lookup values.

6. Lastly, just turn on the scenarios in make.com, so that it can switch into Live mode, and wait for every 15min to pull data from Email. You basically have an AI powered smart customer service which able to help to route, and suggest the reply, so that the customer service admin speed up the customers response work!

So what’s the verdict? Is NocoDB on-par with Zendesk as a customer service helpdesk? I would said No, far from it. Lack of permission control, and sub form features, which make Zendesk a more complete and powerful helpdesk.

However, having said that, NocoDB is still my favorites tool, when come to low code deployment, and POC. If you having a small team of support personnel, I would said NocoDB work just fine.

Although the suggestion from OpenAI GPT-3.5 Turbo can be handy, but it shouldn’t send out directly to the visitors, as I notice sometime the suggestion that OpenAI response are for the admin, not the visitors. In this case, I would suggest that admin should at least check thru the suggestion before sending out to visitor.

For spam, if the message is too vague, OpenAI GPT-3.5 Turbo also can’t determine clearly.

Please take note, OpenAI GPT-3 and OpenAI GPT-3,5 Turbo having some slight differences in the prompt setup, which is OpenAI GPT-3.5 Turbo is designed to be chatbot oriented
Leave a Reply

Your email address will not be published. Required fields are marked *

Prev
OpenAI GPT-3.5 Turbo x Email x make x Zendesk: Low Code Smart Customer Service

OpenAI GPT-3.5 Turbo x Email x make x Zendesk: Low Code Smart Customer Service

Having different scenarios to handle in your customer service?

Next
APITable: Web contact form in 5 minutes

APITable: Web contact form in 5 minutes

Looking for alternative to SeaTable or Airtable for web contact form?

You May Also Like
Total
0
Share