← All case studies
R Ramp · $32B fintech · 3,000+ employees

They set up AI for people. Not the other way around.

The result: 99% daily adoption, underwriting cut from days to hours, and AI baked into how the whole company works.

99.5%
daily AI adoption across the company
2d → 4h
underwriting turnaround
4–5 hrs
reclaimed per employee, per week
The problem

Most employees wanted AI. They just couldn't get past setup.

Like every other company in 2024, Ramp's leadership saw what was happening with AI and pushed adoption. They ran trainings. They bought seats. They shared prompts in Slack.

Six months later, adoption was patchy. The engineers loved it. Everyone else used ChatGPT in private tabs, asked it generic questions, got generic answers back, and decided AI "wasn't for them."

The issue wasn't talent or willingness. It was setup tax. Every employee was being asked to construct their own context layer from scratch — explain Ramp, paste in their data, remember what tools to use. Most gave up before the third try.

What they built

Three layers, stacked into one workspace.

Instead of asking every employee to configure their own AI, Ramp built a stack that did the configuring for them. New hires get it on day one — fully loaded with Ramp's context, voice, and skills.

Ramp's AI stack
Layer 01 · Workspace
Glass
An internal AI workspace pre-configured for every new hire. Pulls memory from Slack, Notion, Linear, and Salesforce. Already knows about Ramp — the company, the voice, the rules — before the employee types a single word.
Layer 02 · Skills
Dojo
A shared library of 350+ "skills" anyone can use. Underwriting, reconciliation, customer-digests, board deck drafts. Each skill is a recipe the AI follows. Lawyers, ops, finance — everyone builds with the same library, and every new skill compounds.
Layer 03 · Agents (e.g. Toby)
Toby
A named AI agent who has read every customer conversation, every NPS comment, every sales call. Quotes them on demand. When customer success needs to understand a churn risk, they ask Toby — and Toby answers like the most senior person in the room.

The key move: no setup tax. A new Ramp employee logs in on day one and Glass already knows who they are, what team they're on, what data they have access to, and what tools they should be using.

The result

99% daily adoption. And it kept growing on its own.

The numbers spoke for themselves. 99% of Ramp employees use AI every single day. Underwriting turnaround dropped from 2 days to 4 hours. Beyond a single workflow, the deeper win was that AI became how Ramp's whole team worked — engineers, ops, finance, customer success — all pulling from the same Glass + Dojo stack.

The cultural shift mattered more than any single metric. AI became the default starting point for any task, instead of a tool you reached for occasionally. The 5-minute Slack message turned into a 30-second Toby query. The 2-hour deck became a 10-minute "draft me the Q3 board update."

We launched our chat interface, called Glass, with no AI training. Everyone just opened it like a doc. — Eric Glyman, CEO, Ramp
What you can steal

Stop teaching your team how to use AI. Set AI up so they don't have to learn.

The thing Ramp got right: they treated setup as the bottleneck, not skill. Most companies invest in training. Ramp invested in a workspace that needs no training. Their employees aren't more "AI-savvy" — they just don't have to be, because Glass + Dojo + Toby do the heavy lifting before they even open a chat window.

If your team is "not using AI enough," the bug is almost never the people. It's the missing context layer. That's the gap we close.

Sources

Eric Glyman — "How Ramp uses AI for internal productivity" (ramp.com)