Generating coaching insights using large language models to improve pulse status

1. A method comprising:

determining, from one or more data sources indicating activity of a user account, a pulse status defining a measure of productivity of the user account relative to a predefined objective of the user account;

generating, based on the pulse status of the user account, a coaching prompt comprising instructions to generate a suggestion for improving the pulse status of the user account;

generating, utilizing a large language model to process the coaching prompt, a coaching insight comprising a recommended action for improving the pulse status of the user account; and

providing the coaching insight for display on a client device associated with the user account.

2. The method of claim 1 , wherein determining the pulse status comprises:

generating a pulse status prompt from observation layer data defining content items presented on the client device and from world state data defining device metrics from sensors of the client device; and

utilizing the large language model to generate the pulse status from the pulse status prompt.

3. The method of claim 1 , further comprising determining the predefined objective of the user account by receiving, from the client device, an indication of a target objective made up of a plurality of executable processes that, when performed, accomplish the target objective.

4. The method of claim 1 , wherein generating the coaching prompt comprises generating text that instructs the large language model, wherein the text includes terms from an observation layer data source and terms from a world state data source.

5. The method of claim 4 , wherein generating the coaching insight comprises utilizing the large language model to generate, from the coaching prompt, text defining the recommended action in a format based on one or more of the observation layer data source or the world state data source.

6. The method of claim 1 , wherein providing the coaching insight for display comprises providing, for display on the client device, a coaching insight interface comprising the coaching insight and one or more cards selectable to redirect the client device to computer applications corresponding to the coaching insight.

7. The method of claim 1 , further comprising modifying parameters of the large language model based on sample executive data defining ground truth coaching insights for sample coaching prompts.

8. A system comprising:

at least one processor; and

a non-transitory computer readable medium comprising instructions that, when executed by the at least one processor, cause the system to:

determine, from an observation layer data source and a world state data source associated with a user account, a pulse status defining a measure of productivity of the user account relative to a predefined objective of the user account;

generate, based on the pulse status of the user account, a coaching prompt comprising instructions to generate a suggestion for improving the pulse status of the user account;

generate, utilizing a large language model to process the coaching prompt, a coaching insight comprising a recommended action for improving the pulse status of the user account; and

provide the coaching insight for display on a client device associated with the user account.

9. The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to:

decompose the predefined objective into a series of executable processes that, when performed, accomplish the predefined objective; and

determine the pulse status by determining a measure of accomplishment of an executable process from among the series of executable processes.

10. The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to:

generate the coaching prompt to include text representing first data from the observation layer data source and second data from the world state data source; and

generate, utilizing the large language model to process the coaching prompt, the coaching insight to intercept erroneous behavior of the user account indicated by the first data and the second data.

11. The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to generate the coaching insight by:

capturing, from the world state data source, a world state moment comprising a state of environmental metrics associated with the client device at a point in time; and

generating a text description of the recommended action corresponding to improving the pulse status in relation to the world state moment.

12. The system of claim 11 , further comprising instructions that, when executed by the at least one processor, cause the system to provide the coaching insight for display by providing a coaching insight interface that depicts the world state moment together with the coaching insight.

13. The system of claim 8 , wherein:

the observation layer data source indicates content items presented in application windows on the client device over time; and

the world state data source indicates device metrics and environmental metrics associated with the client device.

14. The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to generate the coaching insight by generating a suggestion to improve an ongoing interaction associated with the user account based on the coaching prompt indicating observation layer data and world state data for the ongoing interaction.

15. A non-transitory computer readable medium comprising instructions that, when executed by at least one processor, cause the at least one processor to:

determine, from one or more data sources indicating activity of a user account, a pulse status defining a measure of productivity of the user account relative to a predefined objective of the user account;

generate, based on the pulse status of the user account, a coaching prompt comprising instructions to generate a suggestion for improving the pulse status of the user account;

generate, utilizing a large language model to process the coaching prompt, a coaching insight comprising a recommended action for improving the pulse status of the user account; and

provide the coaching insight for display within a coaching insight interface presented on a client device associated with the user account.

16. The non-transitory computer readable medium of claim 15 , further comprising instructions that, when executed by the at least one processor, cause the at least one processor to determine the pulse status by:

determining the measure of productivity in relation to accomplishing an executable process included in a series of executable processes that make up the predefined objective;

comparing the measure of productivity with previous measures of productivity associated with the user account; and

generating the pulse status based on comparing the measure of productivity with the previous measures of productivity.

17. The non-transitory computer readable medium of claim 15 , further comprising instructions that, when executed by the at least one processor, cause the at least one processor to generate the coaching prompt by generating text that includes terms based on an observation layer data source and a world state data source from among the one or more data sources.

18. The non-transitory computer readable medium of claim 15 , further comprising instructions that, when executed by the at least one processor, cause the at least one processor to generate the coaching insight by utilizing the large language model to generate, from the coaching prompt, the coaching insight in a format based on an observation layer data source and a world state data source from among the one or more data sources.

19. The non-transitory computer readable medium of claim 15 , further comprising instructions that, when executed by the at least one processor, cause the at least one processor to generate the coaching insight by:

capturing, from an observation layer data source of the one or more data sources, an observation layer moment comprising a presentation of digital content depicted across a set of application interfaces on the client device at a point in time; and

generating a text description of the recommended action corresponding to improving the pulse status in relation to the observation layer moment.

20. The non-transitory computer readable medium of claim 19 , further comprising instructions that, when executed by the at least one processor, cause the at least one processor to provide the coaching insight for display by providing a coaching insight interface that depicts the observation layer moment together with the coaching insight.