Ramp gives finance teams controls over coding-agent spend

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 1:17:46

   perspective but from the cfo's perspective is the idea that um by being sort of an independent third party as ramp has been in the the big lab race you can design an agent that uses the the best tool for the job. And sometimes that's an OpenAI model. Sometimes that's an anthropic model. Sometimes it's a Gemini model. It might be an open source model. It might be something that's fine-tuned. Is there a value where you believe that you can drive the cost per task lower than the leading labs can provide if you're all in on one? And so this allows a company to be multi-lab? Or do you eventually want to get into inference as well? How deep does that go?

2. 1:18:23

   Yeah, I mean, I think so. The way I think about it is a lot of complexity that these businesses have to solve for is to get their contacts and their data and their repo set up in a lot of different coding agents. And that forces you to effectively pick. And right now, I'm not sure who's winning. And the leading model provider changes very, very quickly every few weeks. And so if you had to pick, I don't know what the answer is. And so we wanted to build something that works across all the tools. And you can always switch where half the company works on models from a certain lab and the other side of the company He's very opinionated and so we have every provider running inference here and so we also run our own inference for InSpec and we're happy to work with customers for that as well.

3. 1:19:06

   Our focus, as you said, is we're not incentivized to just drive up token costs and so for us it's like deeply integrated to ramp router where you understand where every single session is going, which department's using how much, how many tokens of which model, which service tier and which reasoning effort. There's a lot of complexity that we're trying to expose to finance people so they can understand stand and also have controls, in addition to engineers, on how these dollars are being spent. There's an enormous amount of money flowing through, and I mean, finance people have to be involved.

4. 1:19:36

   Howard, I imagine there's like a bunch of other products like this that you guys are using right now as internal tools. Talk about the decision-making process of when you want to open them up to other people through AMP customer base versus when they should just stay internal tools. Yeah, great question. I mean, there's a ton of complexity and automation that actually Inspect sets up for us. And so a lot of...

5. 1:20:00

   Our automations and internal tools actually works on top of InSpec's API. So for us, InSpec is an internal coworker. You tell InSpec, hey, this person said they had a bug or something is slow or something's not working or their support they wrote in or we would like to sell to them, so what should we say to them? All kinds of this job you can offload InSpec because it's plugged into RAMP systems. And so we've also since then set up automations to review code to make sure that code is working correctly when it gets to production.

6. 1:20:28

   Cybersecurity is a big one for InSpec. And so all of these internal automations have now kind of been sitting on top of this. And so for us, it's the bedrock, which sits on top of a router. So we're building these layers of abstractions, and then we can expose to our customers.