UserTesting and Artificial Intelligence

This article provides an overview of UserTesting's artificial intelligence and machine learning (ML) capabilities and how data generated from Artificial Intelligence (AI) is always managed in a secure, compliant manner.

This article applies to: ut logo tiny.pngUserTesting 

On this page:


 

AI Overview

UserTesting has been investing in AI/ML since 2019, optimizing our machine-learning models to not only speed up time to insights but also to help teams focus on more strategic work by surfacing insights that drive customer-centric decision-making across the enterprise.

  • AI powers every stage of the research lifecycle, accelerating everything from the way organizations connect to their customers to post-test analysis and insight summarization.
  • UserTesting's AI capabilities help you pinpoint key insights without spending hours watching videos or connecting the dots across data.
  • AI-generated insights always point to the source data so that you have a way to back them up.
  • Read more about UserTesting's vision for AI on our blog. 

 

How UserTesting uses AI

  • Interactive Path Flows: Visualizations that show how contributors navigate a website or prototype. They generate behavioral data as contributors complete a task.
  • Sentiment Path: An interactive visualization laid on top of the Interactive Path flow that automatically evaluates and summarizes sentiment (positive/negative) feedback from web-based experiences.
  • Intent Path: An interactive visualization laid on top of the Interactive Path Flow that groups specific customer behaviors (e.g. browse, add to cart, search) based on that individual's intent.
  • Keyword Map: An interactive visualization that evaluates verbal tasks and draws out adjectives that contributors used most frequently.
  • Sentiment Analysis: An ML-generated indicator that surfaces moments of negative and positive sentiment when reviewing a completed session in the UserTesting video player.
  • Smart Tags: An ML-generated indicator that highlights themes (e.g. easy, pain point, suggestion) in the video player and for written tasks.
  • Friction Detection: An ML-generated indicator that offers insight into where contributors had difficulty interacting with websites or prototypes during tests.
  • AI Insight Summary: An AI-generated tool that summarizes the tasks and findings captured in Interactive Path Flows and verbal tasks.

 

Data security

UserTesting is dedicated to enterprise-grade information security and the protection of confidential data. We ensure that we comply with applicable law, our contractual agreements, and our rules and policies protecting personally identifiable information (PII) when training our models. 

 

Model training data

  • We use various types of machine learning models including those pre-trained on publicly available data and unsupervised models that don't require training from customer data.
  • Models trained on customer data use data that has been aggregated and anonymized and do not leverage identifiable information to generate results.
  • No unaggregated, confidential data is learned from one account and used for another.

 

Types of data used

  • UserTesting limits the amount of data being used to generate the task summary.
  • Only the task prompt, transcripts for the task, and click data (behavioral data) are used.

 

Open AI

  • UserTesting only shares with OpenAI the data necessary to deliver direct customer benefit such as the data that the customer requests to be summarized through the AI Insight Summary feature.
  • No data processed by OpenAI is used to train their models. The data does not become a part of OpenAI corpus.

Learn more about how UserTesting protects and secures customer data →

 

 

Related content

information icon.png

knowledge icon.png

Want to learn more? Check out these Knowledge Base articles... 

Interested in growing your skills? Check out our University courses...

video icon 2.png

team icon.png

Need hands-on training?

Can't find your answer?

Please provide any feedback you have on this article. Your feedback will be used to improve the article and should take no more than 5 minutes to complete. Article evaluations will remain completely confidential unless you request a follow-up. 

Was this article helpful?
0 out of 0 found this helpful