Finance ops · Production

Ramp customers run month-end close with AI assistants via MCP and CLI

The problem

Finance teams had no standardized way to connect AI assistants to their financial software; every integration required credentials, custom code, and ongoing developer maintenance, making AI-assisted finance work inaccessible to non-developers.

First attempt

Prior to MCP/CLI connectors, connecting AI to financial software required building custom integrations, creating a barrier for finance operators who are not developers.

Workflow diagram · grounded in source
1
Connect AI assistant to Ramp
integration
“MCP and CLI servers are lightweight "connectors" that let an AI assistant (or your terminal) securely talk to Ramp, so you can ask questions or take actions without building an integration yourself”
2
Finance operator submits query
trigger
“Controllers, finance managers, and business owners now run month-end close, reconcile bills, and approve expenses through Claude, ChatGPT, Notion AI, or their terminal”
3
AI retrieves and processes financial data
ai_action
“has spent over 10 hours tuning Claude to generate financial statements, run flux analysis, conduct internal audits, and build one-click reporting”
4
Human retains payment and communication authority
human_review
“He doesn't approve or release payments through AI — that stays in Ramp's product, where there's a full audit trail. He's comfortable letting Claude communicate with vendors, but not yet with customers”
5
Completion-oriented action executed
output
“62% of usage is completion-oriented work like coding transactions, attaching receipts, and approving requests”
Reported outcome

More than 2,500 businesses now connect AI assistants to Ramp for finance operations, growing from 275 at launch, with 62% of usage being completion-oriented actions like coding transactions and approving requests.

Reported metrics
businesses connected to AI assistant2,500+
Businesses connected at launch275
Weekly users pulling spend breakdowns52%
Weekly users searching bills and invoices38%
Show all 15 reported metrics
businesses connected to AI assistant2,500+
businesses connected at launch275
weekly users pulling spend breakdowns52%
weekly users searching bills and invoices38%
weekly users finding missing memos33%
weekly users auditing reimbursements and GL codes14%
weekly users batch-approving transactions12%
weekly users who are business owners60%
weekly users who are business admins36%
completion-oriented usage share62%
finance tools available via MCP/CLI50+
time spent tuning Claude by one controllerover 10 hours
prior AP platform annual costup to $20k a year
AP platform savings versus prior vendorearning back well over that amount
total businesses using Ramp50,000+
Reported stack
MCPRamp CLIClaudeChatGPTNotion AIERP
Source
https://ramp.com/blog/ramp-customers-are-running-month-end-with-ai-assistants
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

More than 2,500 businesses now connect AI assistants to Ramp for finance operations, growing from 275 at launch, with 62% of usage being completion-oriented actions like coding transactions and approving requests.

What tools did this team use?

MCP, Ramp CLI, Claude, ChatGPT, Notion AI, ERP.

What results were reported?

businesses connected to AI assistant: 2,500+; Businesses connected at launch: 275; Weekly users pulling spend breakdowns: 52%; Weekly users searching bills and invoices: 38% (source-reported, not independently verified).

What failed first in this deployment?

Prior to MCP/CLI connectors, connecting AI to financial software required building custom integrations, creating a barrier for finance operators who are not developers.

How is this finance ops AI workflow structured?

Connect AI assistant to Ramp → Finance operator submits query → AI retrieves and processes financial data → Human retains payment and communication authority → Completion-oriented action executed.

Related finance ops cases