An internal incubator at Meta Platforms, Inc. responsible for developing AI products and tools is creating a service to rival OpenRouter. This service aims to reduce costs by directing some AI tasks to run on lower-cost models.
The incubator, known as AAI Labs, falls under Meta's Applied AI Engineering team, which was formed in March of this year. Employees can propose AI projects and services for internal use, with the potential for successful projects to be released publicly in the future. According to a review of internal documents, once a project proposal is approved, a small team is assembled to develop it, with a product launch planned for the opportune moment.
Key Details on the New Service
A July internal memo indicates that AAI Labs has approximately 200 approved projects spanning consumer-facing products, developer tools, and internal infrastructure.
One such project is an AI model routing tool called Switchboard. It functions by scoring the difficulty of a task to determine which model should handle requests from human users or AI agents. Documents show the tool will assign simpler requests to smaller, more cost-effective models, operating in a manner similar to OpenRouter's automatic routing product.
Sources familiar with the matter state that, like many projects within AAI Labs, Switchboard is in the early stages of development and may not necessarily be launched publicly. However, the project proposal indicates the team believes Meta Platforms, Inc. could use Switchboard internally to cut costs or release it publicly for organizations deploying AI code agents at scale.
Strategic Objectives Behind the Projects
The projects incubated by AAI Labs reflect Meta's ambition to transform its massive AI investments into new tools, business segments, and revenue streams beyond its core advertising business. The company forecasts its spending on AI infrastructure, various hardware, and related facilities could reach $145 billion this year, more than double its projected 2025 expenditure. Concurrently, Meta has been restructuring its engineering teams to bolster AI research and development capabilities.
These projects also demonstrate Meta's strategy of leveraging employee-driven innovation to rapidly prototype AI products that can both optimize internal operations and be marketed as standalone offerings.
Focus on Cost Management
Meta Platforms, Inc. has been actively seeking ways to control the billions of dollars spent on procuring AI tools for programmers and other staff. It was previously reported that in June, Meta informed employees it would begin limiting AI token usage, just weeks after encouraging widespread adoption of AI tools across the company. In tandem, Meta is building an internal platform to track AI-related expenditures and enforce token budgeting controls.
Comments