How to process & follow up on customer feedback with Levity

How to process & follow up on customer feedback with Levity

Luc Meijer



Customer feedback is an essential part of any business. Without customer feedback, you are far less likely to understand what drives customer satisfaction. A Harvard Business Review study found that asking customers for feedback leaves a positive impression.

Although customer feedback is extremely important, manually processing it can cost a lot of time. Using Levity and its Zapier integration to process the contents of feedback and Zapier to handle the follow-up can improve your customer feedback process significantly and will allow you to scale efficiently.

In this article, you'll learn how to analyze the sentiment of your customer feedback and act accordingly.

Decisions to make about the automation

Depending on the tooling you use for your business there are three things you might want to adjust for the automation:

1. What triggers the automation

2. A confidence score of sentiment analysis you're comfortable with. We went with 75%

3. What happens in the following cases:

  • Positive customer feedback
  • Negative customer feedback
  • A review that did not match any label (positive or negative) with a confidence score of ≥75%

If you don't have a tight process set up yet, or if you would like to improve it, we'll go through some common scenarios. We'll walk through the options for the following parts of the automation:

  • the trigger
  • positive feedback
  • negative feedback
  • insufficient confidence

What triggers the automation

  • A new row in a Google Sheet containing customer feedback data
  • A new survey comes in from Typeform, JotForm, or any other form tool that collects customer feedback
  • A new review in Shopify, WooCommerce, or any e-commerce tool
  • Tools with specific customer feedback functionalities such as TrustPilot or HubSpot

Positive feedback

  • Ask the customer for a testimonial
  • Push the review to the website
  • Share the review on social media
  • Create a follow-up task in a project management tool like Asana, Monday, or Trello

Negative feedback

  • Notify the customer experience team of the negative review
  • Schedule a task to ask the customer for feedback in the future to see if their experience has improved
  • Create a follow-up task in a project management tool

Insufficient confidence

  • Create a task for the customer experience team to manually process the review

Try it out for yourself with this text analysis widget!

Step-by-step guide

1. Use the Sentiment template. Using this template allows us to work with a pre-trained model that will classify the sentiment of our customer feedback.

Using templates at Levity to get started
Choose the sentiment template

2. In the newly created AI Block, head over to the integrate tab to enable the Zapier integration


3. Go to and create a new Zap

4. Choose any trigger. As we've mentioned earlier, this can be a Google sheet or any other source of customer feedback. Make sure this step includes all the data you need from your customer. For example, their email would be handy!

5. Pick Levity for the next step

6. Take the classify text action

classify text with Levity + Zapier
What do you want your model to do?

7. To use the confidence scores, we'll need to reformat the results. So, for our next action, we'll use Zapier's Formatter. For the action event, pick Utilities.

Setting up Zapier and Levity
Select Utilities

8. When setting up the action, use the Line-item to Text transformation. Then, use the Predictions Results Confidence as the input value. We'll separate this entry by a ,. Your step should look as follows:

Setting up HITL in Levity and Zapier
Don't forget the comma!

9. Although Levity's models are great, they can be wrong sometimes. That's why Levity provides a confidence score for each classification of text. We should base our actions on the confidence score. We went with a confidence score of 75%. Meaning that we'll handle the review differently if the algorithm is less than 75% sure the review is either positive or negative.

10. The next steps will be based on the confidence score of the sentiment analysis. We'll be using the Path app for our next step, and create 3 paths:

  • Positive feedback
  • Negative feedback
  • Insufficient confidence

11. The first path will be for when the review is classified as positive. Output Item 1 contains the score for the positive label. We've settled on a confidence score of 75%, so we'd expect the output to be greater than 0.749.

setting up conditionals for HITL on Levity and Zapier
Setting up the rules for positive cases

12. Replicate the same step for the second path, which will be for the negative cases. Make sure to pick Output Item 2 for this path.

Negative cases or HITL on Levity
Setting up the rules for negative cases

13. In some cases, the confidence score might not be higher than 75% for either of our labels. In such a case, we'd like someone from customer experience to manually process the review. Set up the path as follows

Custom HITL on Levity
How to handle edge cases

14. Now that you've set up the paths for the various cases, you can expand upon each of them. We've mentioned potential follow-ups earlier in the article.

That's it! Having automated this vital part of your business means you get to spend more time on other important things.

Learn more with our ebook, linked below.

This was a guest post from Luc Meijer at Luhhu. An automation agency that helps customer save time, money, and sanity streamlining their processes with Zapier.

Now that you're here

Levity is a tool that allows you to train AI models on images, documents, and text data. You can rebuild manual workflows and connect everything to your existing systems without writing a single line of code.

If you liked this blog post, you'll probably love Levity.

Thank you! Please go to your inbox to confirm your email.
We are sorry - something went wrong. Please try it one more time! In case the problem remains, you can also send us an email to
Sign up

More from our Blog

A Complete Guide to Data Labeling for AI

Data labeling helps teach your AI model what you’re looking for, and can help you save time on your day-to-day business processes.

Read story

AI for Customer Support and Why You Need It

Keeping customers happy is a must—find out how AI could help you take your customer service to the next level.

Read story

AI in Finance: Defining Your Automation Strategy & Use Cases

Learn how to implement AI in the financial sector to structure and use data consistently, accurately, and efficiently.

Read story

Stay inspired

Sign up and get thoughtfully curated content delivered to your inbox.