How We Cut AI Model Costs by 44% | Max Peters
MTS
October 19, 2026
Adapta CEO Max Peters explains how the fully bootstrapped Brazilian company uses Vercel's AI Gateway and custom routing to reduce AI model costs by 44%. He reveals why product-led growth is failing SMBs in 2026 and outlines their expansion strategy.
Max Peters
I don't think PLG is enough. So for our type of customers, which are SMBs, like proudly led growth is not enough. Like you having a good UX is not enough. People need the human component to actually drive the most value out of AI. And that's a big bet we're making.
SPEAKER_2
Hello everyone, and welcome back to MTS. Today, I am here at the Vercel Ship Conference, and I am joined by Max Peters, the CEO of Adapta. So welcome for joining me. Thank you for joining me on MTS.
Max Peters
Thank you. Glad to be here.
SPEAKER_2
Glad to be here. I think Adapta has such an interesting story. You're one of the big customers of Vercel's AI Gateway. You guys have grown a ton as a fully bootstrapped company. I would love to learn a little bit more about the context and origin of Adapta. How did you guys start, and what is the product that you guys are building?
Max Peters
Nice. So we started just right after ChatGPT.
We had a bunch of different companies, and we said, okay, we need to sell everything and go all in into AI.
The gap was, okay, there's this ChatGPT, which is kind of cool, but people are using it to have fun, and do jokes, and write the lyrics of a song, and stuff like that. Okay, we want to use that technology, but for work, for knowledge work. So since the beginning. We've always been in this trajectory of creating an AI platform in which you can do knowledge work better. Right now, I see ourselves as the operating layer for SMBs. So there's a huge gap in like, there's model capabilities, and there's actual AI value you're getting. So like your actual and native AI company, and we bridge this gap. Okay, we're trying to close this gap.
SPEAKER_2
Yeah, you bridge this gap, you close this gap between the model capabilities and what you're actually spending. And one of the big ways that you guys are doing that is through Vercel's AI Gateway. So how do you guys use AI Gateway, and how is it benefiting your customers?
Max Peters
Nice. So if you think about it, our token usage was increasing a lot, and we were always being model agnostic. So whichever model you want to use, you can use all of them inside our platform.
And to do that, we have to have relationships with a bunch of different providers, each own with their own versability like dashboards and everything. So it was a mess. And then Vercel came, we were already customers of Vercel. And they said, okay, we have this new product we're launching.
I think it can help you guys. And in fact, it did. Because we tried a bunch of different things in-house and to make this easier. And it didn't work. And Vercel was the first one that actually worked for us.
SPEAKER_2
Yeah. Tell us a little bit more about the technical details behind Adapta's Intelligent Model Router. So how does that exactly work? You guys had some really interesting stats saying you're able to reduce costs by 44 percent while keeping up the quality. How does that all work?
Max Peters
Yeah. Cool. So what we do, we have this router.
So to choose which model we're going to choose. So think about it, the user makes a prompt. And before we go to the AI gateway and request a model, we have this router which will select the model and base on the prompt of the user. And we do that by creating like internal evals and making sure we understand which models answer better each user request. And since we know really well who we're serving and the use cases what they're doing, we can do that in a really efficient way, in which we can deliver frontier model quality, but with way less cost.
SPEAKER_2
One thing I'm curious about too, is when it comes to AI gateway, there's also different types of routers, there's different type of gateways that do exist. And so for you and your company, how did you guys pick Vercel's AI gateway? What was the big reason for that?
Max Peters
I think one of the reasons is because we already had the relationship, of course, but that makes us try it. What made us stick with it was that it actually worked way better than the other things that we had before. So it was way easier to have all the models inside just one place. So we had relationship with one big, with Vercel instead of having relationships with 30 different model providers and having to use each one have their own API rules and stuff like that. So it was way easier to have everything just with Vercel.
Read the other 3 minutes of this episode here.
$15 gets you this episode and 99 more. Credits never expire. After checkout, one link brings you back here to read it.
Not what you expected? Full refund within 14 days.
Thousands of transcripts fetched by people building searchable podcast archives
Fetch transcripts with code
The same credits work from a script or an AI agent. The demo key returns a sample episode in full, no card needed:
curl -H "x-api-key: pt_demo" \
https://spoken.md/transcripts/1000651996090Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.
From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.
Using your own key:
curl -H "x-api-key: YOUR_KEY" \
https://spoken.md/transcripts/1000797083730