Bugs Fixed:


1) Making AI Features More Reliable and Useful:

This month, we focused on improving the reliability, visibility, and usability of our AI-powered features. Users can now access imported OpenAI assistants without errors, provide feedback on AI evaluation responses, and manage evaluations more easily through a dedicated list view. We also improved evaluation performance for large datasets and introduced a structured sign-up process for teams interested in AI Evaluations.

Highlights:

• Improved monitoring and tracking of AI evaluation performance.

• Fixed issues with imported OpenAI assistants.

• Added user feedback on AI evaluation responses.

• Introduced a dedicated AI Evaluations list view.

• Improved evaluation performance for large datasets.

• Enforced unique evaluation names within a team.

• Added a structured AI Evaluation onboarding process.
  
  
  

2) A more stable voice AI Experience:

Several improvements were made to voice-based AI features to reduce failures and improve reliability. Better error handling, logging, and backup processing mechanisms ensure a smoother experience even during periods of high demand.

Highlights:

• Added backup speech recognition support during high traffic.
• Improved handling of audio processing and file storage failures.
• Enhanced error tracking for voice AI and voice search features.
• Faster issue diagnosis through improved logging and monitoring.
  
  

Enhancement Done:


1) Better WA Form Responses :

We made improvements to WhatsApp interactions to ensure responses are captured correctly and messaging features work as expected. Users can now rely on more accurate form submissions and improved support for WhatsApp group conversations.

Highlights:

• Fixed multiple-choice response capture in WhatsApp Forms.
• Improved support for WhatsApp group memberships.
• Fixed dynamic links in secondary Call-to-Action buttons.

Documentation Update :

New Documentation Added:

1) AI Evaluations :

We introduced AI Evaluations to help organizations measure how accurately their AI Assistants respond to questions. By comparing responses against a trusted set of golden question-and-answer pairs, teams can identify gaps, improve performance, and ensure their Assistants provide reliable information before deployment. Please read more about it here

2) Speech-to-Text Capabilities in Glific:

We introduced Speech-to-Text capabilities in Glific, enabling organizations to automatically transcribe voice notes and translate them into a common language using AI. This makes it easier to collect and understand responses from beneficiaries who communicate in different regional languages.Read more about it here.