IP Library Granted Patent US 12705418
Granted Patent B2
US 12705418 · App. 18/585,995 · Granted Aug 11, 2026

Efficient multi-turn generative AI model suggested message generation

Inventors: Susan Marie Grimshaw (Kirkland, WA); Poonam Ganesh Hattangady (Seattle, WA); Caleb Whitmore (San Francisco, CA); Tashfeen Ahmed (Dublin, IE); Ravi Teja Koganti (Bellevue, WA); Michael Ivan Borysenko (Brighton, CA)
Assignee: Microsoft Technology Licensing, LLC
G06F40/166G06F3/0482G06F3/04847G06F40/205G06F40/40G06N3/0475G06N3/09
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12705418
App. No.
18/585,995
Granted
Aug 11, 2026
Kind
B2
Abstract

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.

Claims (95)

1 . A system for generating a suggested reply message using a generative artificial intelligence (AI) model, comprising:

a processor; and

memory storing instructions that, when executed by the processor, cause the system to:

receive a user-input selection of a message;

combine at least a portion of the message and a predefined request phrase to form a first prompt, the predefined request phrase requesting multiple draft replies;

provide the first prompt to a generative AI model;

receive, in response to the first prompt, a first output from the generative AI model including the draft replies;

causing a display of one or more user interface elements representing the draft replies;

receive a selection of one of the draft replies;

generate a second prompt by combining the portion of the message, data based on the selected draft reply, and additional context for the message;

provide the second prompt to the generative AI model;

receive, in response to the second prompt, a second output from the generative AI model including a suggested draft reply;

cause a display of the suggested draft reply;

display, concurrently with the suggested draft reply, one or more selectable customization user interface elements;

receive a selection of one of the one or more selectable customization user interface elements; and

in response to receiving the selection of the customization user interface element, cause the suggested draft reply to be adjusted by the generative AI model.

2 . The system of claim 1 , wherein the instructions further cause the system to parse the first output to generate a shortened summary for each of the draft replies.

3 . The system of claim 2 , wherein the instructions further cause a display of the shortened summaries as part of the one or more user interface elements representing the draft replies.

4 . The system of claim 2 , wherein the suggested draft reply is displayed concurrently with the display of the shortened summaries.

5 . The system of claim 1 , wherein the instructions further cause the system to:

inject the suggested draft reply into a reply message;

provide a send option concurrently with the reply message;

receive a selection of the send option; and

send the reply message.

6 . The system of claim 1 , wherein the additional context includes additional data regarding at least one of a sender of the message or a recipient of the message.

7 . The system of claim 6 , wherein the instructions further cause the system to:

generate a request phrase including at least a portion of the additional context for the message; and

include the generated request phrase in the second prompt.

8 . The system of claim 1 , wherein the predefined request phrase includes a defined number of draft replies to be generated.

9 . The system of claim 1 , wherein the predefined request phrase includes a length-limiting portion.

10 . The system of claim 1 , wherein the message is an email and the reply message is structured as a reply-all.

11 . The system of claim 1 , wherein causing the suggested draft reply to be adjusted by the generative AI model comprises:

generate a next prompt by combining at least the portion of the message, additional context for the message, and a customization option associated with the selected customization user interface element;

provide the next prompt to the generative AI model;

receive, in response to the next prompt, a next output from the generative AI model including a reframed suggested draft reply; and

cause a display of the reframed suggested draft reply.

12 . The system of claim 1 , wherein the one or more selectable customization user interface elements at least one of:

tone customization options; and

length customization options.

13 . A computer-implemented method for generating a suggested reply message using a generative artificial intelligence (AI) model, the method comprising:

receiving a user-input selection of a message having a body and a header;

extracting the body from the message;

combining the body and a predefined request phrase to form a first prompt, the predefined request phrase requesting multiple draft replies;

providing the first prompt as input to a generative AI model;

receiving, in response to the first prompt, a first output from the generative AI model including the multiple draft replies;

causing a display of one or more user interface element representing the draft replies;

receiving a selection of one of the draft replies;

generating a second prompt by combining the body and data based on the selected draft reply;

providing the second prompt to the generative AI model;

receiving, in response to the second prompt, a second output from the generative AI model including a suggested draft reply;

surfacing the suggested draft reply in a user interface;

displaying, concurrently with the suggested draft reply, one or more selectable customization user interface elements;

receiving a selection of one of the one or more selectable customization user interface elements; and

in response to receiving the selection of the customization user interface element, causing the suggested draft reply to be adjusted by the generative AI model.

14 . The method of claim 13 , further comprising:

parsing the first output to generate a shortened summary for each of the draft replies; and

upon generating the shortened summaries, discarding the first output.

15 . The method of claim 13 , wherein the second prompt also includes additional context for the message that includes additional data regarding at least one of:

a sender of the message;

a recipient of the message; and

historical sent message from the recipient of the message.

16 . The method of claim 13 , wherein the predefined request phrase includes at least one of:

a defined number of draft replies to be generated; and

a length-limiting portion.

17 . The method of claim 13 , further comprising:

injecting the suggested draft reply into a reply message;

providing a send option concurrently with the reply message;

receiving a selection of the send option; and

sending the reply message.

18 . The method of claim 13 , causing the suggested draft reply to be adjusted by the generative AI model comprises:

generating a next prompt by combining at least a portion of the message, additional context for the message, and data associated with the selected customization user interface element;

providing the next prompt to the generative AI model;

receiving, in response to the next prompt, a next output from the generative AI model including a reframed suggested draft reply; and

causing a display of the reframed suggested draft reply.

19 . A computer-implemented method for generating a suggested reply message using a generative artificial intelligence (AI) model, the method comprising:

causing a display of a user interface of a mobile-device email application;

receiving a user-input selection of an email from within the user interface;

causing a display of the selected email;

combining the email and a predefined request phrase to form a first prompt, the predefined request phrase requesting multiple draft replies to the email;

providing the first prompt as input to a generative AI model;

receiving, in response to the first prompt, a first output from the generative AI model including the multiple draft replies;

causing a display of the draft replies in the user interface and concurrently with the selected email;

receiving a selection of one of the draft replies;

generating a second prompt by combining the email, data based on the selected draft reply, and additional context for the message;

providing the second prompt to the generative AI model;

receiving, in response to the second prompt, a second output from the generative AI model including a suggested draft reply;

causing a display of the suggested draft reply in the user interface;

displaying, concurrently with the suggested draft reply, one or more selectable customization user interface elements;

receiving a selection of one of the one or more selectable customization user interface elements; and

in response to receiving the selection of the customization user interface element, causing the suggested draft reply to be adjusted by the generative AI model.

20 . The method of claim 19 , wherein causing the suggested draft reply to be adjusted by the generative AI model comprises:

generating a next prompt by combining at least a portion of the message, the additional context for the message, and data associated with the selected customization user interface element;

providing the next prompt to the generative AI model;

receiving, in response to the next prompt, a next output from the generative AI model including a reframed suggested draft reply; and

causing a display of the reframed suggested draft reply in the user interface.