Keep pulling the thread on Geoff Charles.
As of the recording, 50% of Ramp's code is written by AI, an increase from 30% in December.
Geoff Charles predicts that 80% of Ramp's code will be built by AI by March.
In the last year, Ramp shipped over 500 features and surpassed $1 billion in revenue with a team of approximately 25 product managers.
At Ramp, when a product escalation occurs, an AI agent performs the initial analysis, identifies the relevant code, creates a pull request for a fix, and often ships the change.
Ramp has a mandatory requirement for all new hires, regardless of role, to be proficient with modern AI tools.
Geoff Charles believes the traditional role of management in technology companies is "probably dead."
Geoff Charles predicts that the traditional software model is "dead" and will be replaced by AI-powered "co-workers" or agents.
Ramp has an automated pull request review process that approves a double-digit percentage of PRs without human intervention.
Ramp has a customer base of over 50,000 businesses and more than one million end users.
Ramp developed an internal AI agent called "voice of the customer" that analyzes data from Gong, Salesforce, support tickets, chats, emails, and its Snowflake database to answer product questions.
Ramp's "voice of the customer" AI agent can perform analysis in approximately eight minutes that would take a human analyst eight days to complete.
Ramp has transitioned its internal data analysis tooling to a system based on the Snowflake CLI and Anthropic's Claude Code, enhanced with a proprietary database of custom skills.