Most AI agents are still treated like software tools. A person opens an application, gives an instruction and waits for a result.
Bounty is testing a more ambitious model: what happens when the software finds the work itself?
The New York startup has raised more than $1 million to develop a marketplace where clients post paid assignments and autonomous AI agents compete to complete them. a16z speedrun invested $1 million and selected the company for its highly competitive Speedrun accelerator. Additional backing from close associates pushed the reported total above $1 million, with a16z remaining its only disclosed institutional investor.
The platform resembles a freelance marketplace, but the supply side is software rather than human contractors. A client defines a task, sets the payment and lists the conditions the finished work must satisfy. Agents identify suitable assignments, and the highest-rated eligible agent can claim the job. Bounty holds the payment until the client approves the result, then releases it to the agent’s owner while retaining a 10% commission.
Bounty was founded in early 2026 by Rayan Dabbagh, Aashna Doshi and Shivam Patel. Dabbagh, a 23-year-old Lebanese engineer, previously worked on cloud infrastructure and AI projects at Amazon Web Services. Doshi is a former Google engineer who co-leads the business, while Patel, previously an algorithmic trader at Barclays, serves as chief technology officer.
The founders are not simply building another interface for assigning tasks to a chatbot. They are attempting to create a reputation and payment system for a future in which specialized agents operate as small digital businesses—finding work, building performance histories and generating revenue for their owners.
Early usage remains modest but notable for a newly launched platform. Bounty had recorded more than 1,300 posted tasks by early July, according to Dabbagh. He also said some agents were already earning as much as $500 per month for their owners, although those figures are company-reported and have not been independently validated.
The larger question is whether agent marketplaces can develop the trust mechanisms that human freelance platforms spent years building. Clients need reliable work, clear accountability and a practical dispute process. Agent owners need fair access to assignments and confidence that successful performance will translate into future demand.
Bounty’s early model provides the basic pieces: ratings, escrow-style payment, task requirements and refunds for disputes judged valid. What remains unproven is whether autonomous agents can consistently deliver work valuable enough to support a durable marketplace rather than a short-lived experiment.
The startup’s most interesting idea is not that AI can complete a task. It is that software may eventually participate in an economy—competing, earning and subcontracting work with increasingly limited human direction.
Source : Mena Startup Digit
