Multi-turn Generative AI Message Reply Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing productivity applications require significant time and effort to generate responses to messages due to the need to read and understand the context of previous messages, leading to inefficiencies in content creation.
Innovation Solution
A multi-turn generative AI model is used to generate suggested reply messages, where a shorter prompt is initially used to produce shortened summaries, and upon user selection, a more robust prompt is generated for a second turn to produce a complex suggested draft reply, optimizing the response generation process without sending large initial prompts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large prompt is sent at the outset to generate initial reply options, then more accurate and comprehensive results are obtained, but processing time and computational resources increase significantly
Solution Approach 1:
The prompt processing is divided into two segments: a first turn with a shorter prompt that generates multiple draft replies, and a second turn with a more robust prompt that generates a refined suggested draft reply. This segmentation allows the system to obtain comprehensive results without requiring all processing to occur at once, thereby reducing initial processing time while maintaining accuracy.
Solution Approach 2:
The system performs preliminary action by generating multiple draft replies in the first turn before generating the final suggested draft reply in the second turn. This preliminary generation of options allows the system to prepare multiple potential responses in advance, enabling faster and more accurate final output when the robust prompt is processed.
2Measurement precision
If a large prompt is sent at the outset to generate initial reply options, then more accurate and comprehensive results are obtained, but computational resources are consumed excessively
Solution Approach 1:
The computational workload is segmented across two turns: the first turn processes a shorter prompt to generate multiple drafts, and the second turn processes a more robust prompt to generate the final suggested reply. This segmentation distributes computational resources over time, preventing excessive resource consumption in a single operation while maintaining comprehensive processing.
Solution Approach 2:
The system performs partial action in the first turn by generating multiple draft replies without completing the full processing of a robust prompt. This partial processing reduces immediate computational resource requirements, and the remaining processing is completed in the second turn when resources can be allocated more efficiently.
3Speed
If multiple draft replies are generated and shortened to summaries, then reply options are presented quickly to the user, but detail and complexity are reduced
Solution Approach 1:
The system segments the information presentation into two stages: shortened summaries are presented in the first turn for quick user review, and the full detailed suggested draft reply is generated in the second turn. This segmentation allows users to quickly assess multiple options without being overwhelmed by detailed information, while still providing access to complete details when needed.
Solution Approach 2:
The shortened summaries act as intermediaries between the full draft replies and the user. These summaries provide a condensed representation that enables quick user evaluation, while the system retains the ability to generate and present the full detailed replies in the second turn, thus maintaining information completeness while improving user interaction speed.
Data Source
AI summary
Systems and methods for using a generative artificial intelligence (AI) model using a multi-turn process to generate a suggested draft reply to a selected message. A first turn of the multi-turn process uses a shorter prompt including at least a portion of the body of the selected message and that requests multiple draft replies from the AI model. The resulting AI-generated draft replies are shortened, summarized, and/or otherwise converted into a plurality of shortened summaries that are presented as reply options to a user. Upon selecting a shortened summary, a more robust prompt is generated in a second turn with the AI model with the selected reply option to generate a more complex suggested draft reply to the selected message. Additionally, various customization options are provided, which when selected, reframe a query presented to the AI model to generate a more relevant and personalized response.