Katya: So, I was reading this thing about Ramp, the finance company, and they said that twelve percent of all the human-initiated pull requests on their production code base, it's coming from non-engineers.
Yusuf: Non-engineers? Like, people in sales, in marketing?
Katya: Exactly. Sales, marketing, finance. Thousands of code changes every month. I thought, 'how?'
Yusuf: Thousands? That is a lot. So, what were they doing? Just giving them access to ChatGPT and saying 'go for it'?
Katya: That's what I thought, too! But no, it's much more than that. It's like... they realized that just giving people the tools wasn't enough.
Yusuf: This is what I found interesting, Katya. They actually got 99% of their employees to use AI tools. Like, almost everyone. And then they saw the problem.
Katya: They were stuck, right? Like, they had the car, but didn't know how to drive it fast.
Yusuf: Exactly! They called it like driving a Ferrari with the handbrake on. People had these powerful AI models, but they didn't know how to set them up better, how to improve their prompts, or even share what they learned with anyone else.
Katya: So, it's not about the model itself, it's about the... environment around it. The way you use it.
Yusuf: Precisely. The main barrier wasn't the AI's power, but the friction in the user's environment. They said the goal shouldn't be to simplify AI by removing options, but to make the complexity invisible while keeping all the power.
Katya: That's a very clever way to put it. Invisible complexity.
Yusuf: Right? And it made me think, this is exactly what happened with my friend's marketing agency and Canva. They gave everyone a Pro account, unlimited access to all the fancy features.
Katya: Okay, sounds good, no?
Yusuf: It sounds great. But only two people out of thirty actually learned how to use the 'Brand Kit' feature. So, all these client presentations, you'd see the logo slightly off, the brand colors just a little bit wrong.
Katya: Oh, no. This is the worst.
Yusuf: It was. The value wasn't in having the tool, but in having a system to use it correctly, which they never built. Ramp, on the other hand, they built their own platform they call 'Glass'.
Katya: So, how does Glass work to fix this Ferrari with the handbrake problem?
Yusuf: Well, one big part is something called 'Dojo'. It's a shared skills marketplace. Imagine someone on the sales team figures out the absolute best prompt for analyzing a Gong call, you know, those sales call recordings?
Katya: Yes, I know these. They can be... long.
Yusuf: Very long. So, they package that prompt as a 'skill'. And then, that skill is immediately available to every other sales rep. Instantly, everyone on the sales team is leveled up.
Katya: Like a template, but for AI.
Yusuf: More than a template, it's like a pre-programmed workflow. Then, they have this persistent memory system. It connects to your calendar, Slack, Notion. So, the AI knows your projects, your colleagues. It even runs a nightly process to synthesize your recent work.
Katya: So you don't have to explain everything from scratch every time you talk to it?
Yusuf: Exactly. No re-explaining context. It knows. And then there are scheduled automations. A finance lead, for example, set up a simple prompt to pull the previous day's spending anomalies. Every morning at 8 am, it posts a summary to the team's Slack channel.
Katya: That's... quite smart. No more manually checking reports.
Yusuf: And finally, a multi-pane workspace. Like a code editor, but for everyone. You can draft a Slack message in one pane, review a spreadsheet in another, read a PDF in a third. All in the same window, no constant app switching.
Katya: This sounds... actually very good. It's like they thought about how people actually work, not just, 'here, have an AI chat bot'.
Yusuf: Precisely. They made the tools adaptable to the human, not the other way around.
Katya: Okay, so how do you even get a company to this point? Because this sounds like a lot of effort. This other article I saw, it was talking about how Ramp basically 'AI pilled' their whole company.
Yusuf: Ah, yes, 'AI pilled'. It sounds intense.
Katya: It says you can drive massive adoption without some grand strategy, by just making it a non-negotiable cultural expectation. And creating a ladder for employees to climb, you know, L0 to L3.
Yusuf: A ladder, like... 'if you're still at L0, using ChatGPT sometimes, you probably won't be here long'?
Katya: Basically. It's quite direct. They said their culture is velocity, and that became the biggest accelerant.
Yusuf: Okay, completely different direction here—
Katya: Yeah, I mean, I'm listening to all this and I'm thinking, is this just a massive humblebrag from a hyper-growth, VC-funded tech company? Is it even applicable to anyone else?
Yusuf: But why not? The principles seem sound.
Katya: But a normal company, they cannot afford to build their own AI platform, no? And an 'unlimited budget' to explore? This feels like a playbook for the zero point one percent.
Yusuf: I see your point about the resources, but the core idea... making one person's breakthrough everyone's baseline. That's a mindset. That doesn't necessarily cost millions.
Katya: But it costs development time, it costs people. They had a team of four build 'Glass' in under three months! Most companies don't have that kind of internal capacity to just spin up custom platforms.
Yusuf: True, but the outcome, Katya. A risk analyst automated 16 hours a month. A finance person built a contract reviewer saving 45 minutes per contract. These are massive efficiencies, those pay for the development over time.
Katya: It's easy to say when you're already flush with VC money and have a highly technical workforce to begin with. What about a manufacturing company, or a healthcare provider? Are they just supposed to 'AI pill' everyone and build their own 'Glass'?
Yusuf: I think the argument is that these are the companies that will lead. The ones who realize AI isn't just another tool, it's a fundamental shift in how work gets done. And maybe, in a few years, there will be off-the-shelf solutions that provide a similar integrated experience.
Katya: Maybe. But right now, it feels like they are saying 'just do what we did' when 'what we did' is out of reach for ninety-nine percent of businesses. I'm not convinced you can just mandate this kind of cultural change around AI, unless you already have these hyper-competitive, tech-savvy employees to begin with, no?
Yusuf: Perhaps not universally, but they did say that the early converts mattered most. You find those few people on every team who are curious, who want to build, and you make them visible. That creates a competitive dynamic.
Katya: So, it's a bit like... peer pressure, but for AI adoption? I'm Katya.
Yusuf: That's one way to look at it, yes. And I'm Yusuf. This has been Manish Chiniwalar's Station.