This was v good!

What a great approach. Hugely bullish for your org and product @geoffintech

Very interesting to see how you guys evolved from using Notion to building your own Claude Cowork, but companies without Ramp’s eng team capacity probably don’t want to take on the latter. I think this is partially why even for those few that got through the hump to get on Claude Code (outside of Eng teams for non-coding work), the collaboration surface remains an issue:

👏👏👏 @geoffintech! The number of companies that have asked me about how Ramp has become so AI-native… just sending them this.

How much is ramp spending on tokens per months split between frontier and opensource?

so good

Great piece. You all would love Claude Code on the Roam Map!

This is exactly why big tech Saas is down - they cant do any of this.

You guys are getting close to the cutting edge (ie AI-native startup). Incredible for a company that size! Fun test: what % of employees update >1 markdown file in GitHub daily?

What is your token cost per user for power users?

Super helpful insights in here. Thanks Geoff!

We ran 5 hackathons internally and the more tools & data, the teams done a lot more. I like the hub & spoke strategy right now, on the product side I’m creating environments on @Replit with skill files with our product design & data model. Teams can clone & build on top of.

Thank you for these amazing insights into your process. As we’ve had more enterprise interest in our platform, this has given me a ton of deployment and tools ideas. Really appreciate it.

Interesting. How do you guys handle the engineers time between work and helping others? Do bottlenecks happen of employees who are waiting on an engineer to help them figure something out or a final polish/secure check before hitting prod? Thanks

this is the right framing. most companies think ai adoption is a tooling problem when it’s really an operating system problem. remove setup friction, connect the systems people already use, and get them to a real aha moment fast. that’s when it compounds.

Not included but important to call out, the best way is to have the MOST AI pilled CPO & CTO in tech drive the change everyday.

man, this is a great article. And aligns with my thoughts so much. Love this part "If your internal tools from three months ago still feel state-of-the-art, you're not moving boldly enough." I experienced this when building a tool at amazon late 2024. By May the tool was obsolete but I was like that's fine. AI is moving so fast. that's the new way. Especially we had delivered the app with AI assistance in 4 months. And by may we could develop the same app in a month or two.

Great write up. How do you handle customer PII going into the models? This is usually the main constraint from a security / legal perspective

very interesting to see how you did this at such a big company. also sounds mega stressful to be an employee at Ramp lol

Question for the ramp folks- when you say everyone had connectors already to things like snowflake and SFDC. Is this a read only connector to pull information or does every employee get write access to make changes to these systems via ramp glass?

Curious what tools (if any) you’re using to measure usage? Do you find high level token output metrics useful? Or was sessions/skills more predictive of adoption?

What business goals / measurements can you attibute to the adoption of AI?

I've been following Ramp for some time now... The insight everyone will miss: this isn't a tech story. It's an org design story. Remove constraints, measure publicly, let people build. The tools are the easy part. The culture is the moat. One thing I'd add — this playbook isn't just for operators. If you're an investor, the L0-L3 proficiency ladder is a value creation framework for every portfolio company you touch. The firms that figure that out first will compound faster than anyone running AI as a back-office experiment.

We did an ai guild at servicenow and it had been a powerful force. Their job was to learn-teach-learn ai to the org .

@geoffintech , thanks for sharing this. Really loved how this is down to earth, concrete and coherent road map, that can help people get started. I shared it with our team, portfolio and advisors at @StageOneVC , much appreciated!

This is inspiring. Makes me want to join Ramp!

Love this man so much alignement with how we see things internally too it’s crazy!

Thanks for sharing. Curious - How much has this moved revenue per employee at Ramp?

Is the token usage ROI sustainable?

公司如何进行 AI 转型

Truths. Thanks for sharing.

So good

You really need to check out @qualabi_app

Inspiring.

Do you think the age demographic of Ramp matters? At our company, I'm the AI Adoption Lead and it's been a struggle to get older folks to try it beyond chat, nevermind anything to do with a terminal.

this was such a dank read