How to build a CapEx Classification AI Agent
This AI agent analyzes project documents against CapEx guidelines, verifies the analysis for accuracy, and automatically logs the results to Airtable.
Challenge
Ensuring that project proposals are rigorously and consistently evaluated against Capital Expenditure (CapEx) classification guidelines is challenging, complex, and time-consuming.
Industry
Finance
Department
Finance
Integrations
OpenAI
Airtable
Gmail
Workflow Overview
1. User Provides Inputs
Project/Asset Name:
The user enters the name of the project or asset to be classified (e.g., "Server Upgrades").Reviewer Email:
The user enters the email address of the reviewer who should receive the final report.
2. Upload Supporting Documents
The user uploads all relevant supporting files (invoices, contracts, specs, business case, etc.).
These documents are processed (including OCR for scanned files) so their content can be used by the AI.
3. Knowledge Base Search
Search Knowledge Bases:
The workflow uses the Search Knowledge Bases action to look up relevant information in a specified knowledge base, using the project/asset name and supporting documents as the query.
This helps ensure the AI’s classification is aligned with company policy and best practices.
4. AI Classification
Capex Classification LLM:
The AI reviews the project/asset name, the uploaded documents, and the knowledge base search results.It classifies the project as either Capex or Opex, provides a clear rationale, and cites supporting evidence from the documents and knowledge base.
5. Output and Reporting
Output Node:
The AI’s Capex/Opex classification report is made available for download or review.Send Email:
The final report is automatically sent to the reviewer’s email address provided by the user.
6. Optional: Log to Smartsheet
Prepare Data for Logging:
A secondary AI node prepares the classification data in a format suitable for logging.Add Rows to Sheet:
The workflow can add a new row to a Smartsheet, recording the classification result for tracking and audit purposes.
Node Map (Summary Table)
Node Name | Purpose/Function |
|---|---|
Project/Asset Name | User input for the project or asset to classify |
Reviewer Email | User input for the reviewer’s email address |
Supporting Documents | Upload and process all relevant supporting files |
Search Knowledge Bases | Looks up relevant policy/guidance in the knowledge base |
Capex Classification LLM | AI reviews all inputs and classifies as Capex or Opex with rationale and cites |
Output | Presents the final report for download/review |
Send Email | Emails the final report to the reviewer |
Prepare Data for Sheet | (LLM) Formats the result for Smartsheet logging |
Add Rows to Sheet | Logs the result in a Smartsheet for record-keeping |
How the Data Flows
User provides project name, reviewer email, and uploads documents.
Knowledge base is searched for relevant policy/guidance.
AI reviews all information and generates a classification report.
Report is output for review and sent via email.
Optionally, the result is logged to Smartsheet for tracking.
This workflow ensures a robust, auditable, and automated Capex/Opex classification process, combining user input, document analysis, knowledge base reference, and automated reporting.
