Winona Rajamohan
Product Marketing
We built Seel to close the gaps that drain revenue after checkout. Returns. Reverse logistics. Claims. Support cost. These are the leaks every merchant already knows to watch for.
A new one has opened up as business operators — including ourselves and our customers — put AI agents and tools to drive up productivity.
AI spend.
We're not the only ones seeing this. Data shows that 98% of FinOps practitioners now manage AI spend, up from just 31% two years ago. Visibility into AI costs was one of the top challenges they named.
For merchants, the cost of AI doesn't announce itself the way a spike in returns does. It shows up model by model, provider by provider, request by request, until the bill lands at scaled pricing you didn’t see coming.
Addressing a growing AI concern
We recently announced the launch of Seel Agencies, an operational suite of AI services we’ve opened up to merchants after building them to run our own workflows internally. While developing these systems, we’ve learned that running AI at scale forces you to think about ways to operate with models more efficiently.
We hear the same hesitation in our own merchant conversations. Not a doubt that AI works, but real uncertainty about whether it's worth what it costs, how to prove its value, and what are the best ways to make AI spend feel within control.
When building our model supply infrastructure, our focus was finding a solution that helped us achieve scale, reliability, access, and predictable unit economics. This allows us to run solutions, like our Support Agency, to serve millions of shoppers and actually resolve merchant tickets with a coordinated layer of AI agents.
Solving that for ourselves is what led us to open up our sourcing infrastructure.
What is the LLM Gateway?
The Seel LLM Gateway is an OpenAI-compatible API that gives you access to 25+ frontier and open models through one endpoint. You get your favorite models at discounted pricing with no markup on tokens. spend caps that you can set per key, and automatic failover, so if one provider has an outage, your requests move to another path instead of stalling.
If you're already running on an OpenAI-compatible setup, moving over is a base URL change, not a rebuild.
We’ve also designed our gateway so the model you call is the model that runs and nothing gets swapped out behind the scenes. You get the same sourcing and cost efficiency, without giving up the choice of which model handles your request.
This choice matters. When a model gets swapped mid-production, it can change how an agent reasons, formats a response, or decides when to call a tool, all without anyone noticing the change happened. For workflows like customer support and claims resolution, we believe that control is important.
AI spend is now an operating line
Our mission hasn't changed. We help merchants sell more, save more, and spend less. Support cost, reverse logistics, return risk. These are the categories we already help merchants manage.
AI spend is the newest one on that list. Not a side project or an experiment budget, but an operating cost like any other, one that deserves the same discipline we've always applied to returns and claims.
The Gateway applies that same lens to model spend.
If you want in, request access at seel.com/llm-gateway.
