Spreadsheet LLM Task Decomposition for Low-Latency Responses

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing spreadsheet AI models are limited to spreadsheet data and functionalities, while large language models (LLMs) introduce latency, require significant compute power, and may generate irrelevant or inaccurate responses due to hallucination.

Innovation Solution

An application service integrates LLMs with spreadsheet applications by generating prompts based on user inputs, refining them to focus on spreadsheet capabilities, and receiving clarifying questions to improve response relevance and accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If LLMs are integrated with spreadsheet applications, then the capability to handle diverse data domains and generate novel responses is improved, but latency increases and compute power requirements increase

Engineering Contradiction:

Improvecapability to handle diverse data domainsVSAvoidlatency

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The system segments the LLM integration into multiple components: a prompt generation module that prepares context-specific prompts, an LLM service that processes prompts, and a response generation module that formats outputs. This segmentation allows each component to be optimized independently and enables parallel processing of multiple requests, reducing overall latency while maintaining versatile capabilities.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system performs preliminary actions by pre-processing spreadsheet data into context-specific prompts before LLM processing. The prompt generation module prepares structured queries with relevant context in advance, so when user requests come in, the LLM receives pre-formatted input immediately, reducing processing time and latency.

Inventive Principle:

Principle #10Preliminary action

2Adaptability or versatility

If LLMs are integrated with spreadsheet applications, then the capability to handle diverse data domains is improved, but compute power requirements increase

Engineering Contradiction:

Improvecapability to handle diverse data domainsVSAvoidcompute power requirements

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The system applies local quality by customizing prompts for specific spreadsheet contexts rather than using generic LLM queries. Each prompt is tailored to the particular data structure and user needs, allowing the LLM to process only the necessary information locally, reducing overall compute power requirements while maintaining domain-specific accuracy.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system uses partial action by sending only the necessary portion of spreadsheet data to the LLM through context-specific prompts, rather than transmitting the entire dataset. This selective data transmission reduces compute power requirements while maintaining sufficient context for accurate responses.

Inventive Principle:

Principle #16Partial or excessive action

3Adaptability or versatility

If LLMs are used to process user inquiries, then the capability to generate novel responses is improved, but response relevance decreases due to hallucination

Engineering Contradiction:

Improvecapability to generate novel responsesVSAvoidresponse relevance

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The system implements feedback mechanisms where the LLM service receives context-specific prompts and returns structured responses that are validated against the original user inquiry and spreadsheet context. The response generation module checks for relevance and can request clarifications or adjustments, ensuring high response relevance while maintaining the ability to generate novel insights.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The context-specific prompts act as intermediaries between the LLM and the user inquiry. These prompts translate user requests into precisely formulated questions that reference specific spreadsheet contexts, ensuring the LLM generates relevant responses grounded in actual data rather than hallucinating information.

Inventive Principle:

Principle #24Intermediary (Mediator)

4Adaptability or versatility

If LLMs are used to process user inquiries, then the capability to generate novel responses is improved, but the response may diverge from user intent

Engineering Contradiction:

Improvecapability to generate novel responsesVSAvoidresponse alignment with user intent

Core Design Contradiction:

Adaptability or versatilityVSEase of operation

Solution Approach 1:

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS12536387B2Task decomposition for LLM integrations with spreadsheet environments

Publication Date: 2026.01.27 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

Technology is disclosed herein for the integration of spreadsheet environments with LLM services. In an implementation, an application service receives a natural language input from a user associated with a spreadsheet hosted by a spreadsheet application. The application service generates a prompt based on the natural language input which includes asking a large language model (LLM) service to classify a statement in the input as referring to one of multiple capabilities of the spreadsheet application. The application service inputs the prompt to the LLM service and receives an output from the LLM service which identifies a determined one of the multiple capabilities. The application service generates a revised prompt based on the input and the determined one of the multiple capabilities and inputs the revised prompt to the LLM service.