Preparing and Optimizing Content for AI Training

January 2024-April 2024

 

Uber Metrics-Padding

When our enterprise client began training a generative AI assistant on their existing Help Center content, they quickly learned the generated articles were derived from unhelpful or unreliable sources that were negatively impacting support costs and creating a poor experience for customers.

Steyer’s content specialists were brought in to guide the client in the consolidation, simplification, and alignment of Help Center content for use in genAI-powered features. We used a multifaceted approach to ensure the source documents were as squeaky clean as possible, with standardized formatting, on-brand language, and messaging consistency, using an organized workflow that provided an efficient means of delivering work back to Uber within 24-hour timeframes.

Steyer helps our clients drive business results with trusted content.

Read more about our approach on this project:

A realignment strategy to support better usability, brand language, and consistency of formatting for AI-generated content.

Steyer’s content strategists and writers worked with our client’s editorial team to review the original content given to the AI against the AI-generated output. The team identified areas the AI output seemed to be struggling with, and found opportunities to bring consistency and brand voice back into the client’s Help Center content.

  • The team performed a comprehensive audit of the content using the client’s Tone of Voice and Style Guide documents. This allowed the team to quickly spot words, phrases, and rich text changes made by AI that were not aligned with the client’s style.
  • Flagged duplicate articles and redundancies across articles. This allowed the client to reduce the number of articles in the Help Center, thus reducing time spent searching for answers by customers and Support Agents; it also reduced the chances of a customer or Support Agent encountering conflicting guidance.
  • Identified areas to improve content structure, consistency, and readability. With the client’s “Global-first” mindset at the forefront, Steyer searched for areas to improve the content beyond merely optimizing it for AI training.
  • Steyer ultimately lightened the load for our client’s teams. Steyer’s support not only minimized delays but improved timelines for the client altogether, and resulted in a library of source material that was ready to use in AI training.

We'd love to hear from you!

Thank you. We will be in touch soon!

There was an error with the form. Please check and try again.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.