Highlights:

  • Advanced AI models like GPT-4o can handle various tasks in third-party services, provided those tasks don’t require processing payments.
  • The company claims its platform can connect to AI models and their paid services within minutes, offering a software development kit to simplify the integration.

Recently, a startup that provides software facilitating AI models to make purchases, Skyfire Systems Inc., launched with USD 8.5 million in initial funding.

The investment attracted participation from over a dozen institutional investors, including Ripple Labs Inc., a payment processing software provider backed by over USD 200 million in venture capital. Skyfire’s Co-founders, Amir Sarhangi and Craig DeWitt, both previously held executive positions at Ripple before launching the startup.

Advanced AI models like GPT-4o can handle various tasks in third-party services, but they can’t make payments. Neural networks are currently unable to complete user purchases. Skyfire’s cloud platform wants to fill this gap.

The platform enables AI developers to set up a digital wallet for a large language model. They can fund this wallet using a bank account or USDC, a stablecoin tied to the U.S. dollar. Circle Internet Financial LLC, the creator of USDC, was also involved in Skyfire’s recently announced funding round.

The company’s AI payments platform generates an account for each neural network registered by developers. This account enables the model to access third-party services where it will make purchases. According to Skyfire, the platform maintains an activity log of transactions made by the AI model, which can be used to “demonstrate your agent is a good actor” to sellers.

Users of an LLM can set budget limits to prevent overspending. If the model tries to make a transaction that exceeds its available funds, Skyfire will alert the affected customers.

“Our research indicates the market size of agent to agent commerce alone can reach USD 46 billion over the next three years. We see a future where millions of AI agents transact globally and autonomously through Skyfire,” said Sarhangi, Skyfire’s Chief Executive.

One potential use case for the platform is enabling AI models to make purchases on behalf of consumers. While chatbots like ChatGPT can assist with finding online stores that carry specific products, they can’t complete the purchase. Skyfire allows developers to integrate this capability into their AI services without needing to build a custom payment system.

In some cases, an AI can only answer user questions by accessing information from a paid data source. For example, a financial analyst could request that a chatbot retrieve stock information from a commercial market intelligence feed. According to Skyfire, its platform enables enterprise AI models to purchase access to subscription-based data automatically.

Neural networks using Skyfire can also acquire infrastructure. For example, if a developer requests a coding assistant to run a program that’s too complex for local execution, the neural network could rent a cloud instance and deploy the program on that server.

The company claims that its platform can be linked to AI models and their paid services within minutes. It provides a software development kit to streamline the integration process.

Before its recent launch, Skyfire is said to have built an initial customer base that includes car parts supplier Denso Corp. and several other companies. Moving forward, the software maker intends to expand its platform by incorporating support for additional types of transactions.