Two Google alumni have raised a $11.3 million fund to support early-stage AI startups. They believe the era of enterprise AI experimentation is coming to an end, and customers will increasingly pay only for products that can prove their value.
BAG Ventures was founded by former Google vice president Bontia Stewart and former CapitalG partner Jackson Georges Jr. The fund officially completed its fundraising after approximately two years of preparation. The company has already invested in ten businesses, including the software company SXD, the AI travel agent BizTrip, and the agentic reasoning platform Nomadic.
The fund invests in startups across AI infrastructure, computing power, physical AI and edge AI, security, governance, and vertical SaaS. Investment sizes range from $100,000 to $500,000 per deal, and the team aims to deploy the remaining capital over the next two years.
Stewart worked at Google for 17 years, including nearly a decade as a vice president. During that time, she also served on the board of Gradient Ventures, Google’s early-stage AI fund. She is a limited partner in Female Founders Fund and Operator Collective. She co-led the angel investment syndicate BAG Collective with Georges, which has over 450 members.
George previously worked at GE Healthcare and Google, where he met Stewart. He later became a partner at CapitalG, Alphabet’s growth fund. Both George and Stewart were among the first participants in the Berkeley Black Venture Institute.
Both indicated that their strength lies in "access channels." Georges said they founded BAG Ventures to "bridge the emerging AI gap between founders and operators."
“The founders need to get inside the organizations they want to sell to, and we know many senior operators who are eager to support early-stage founders but aren’t sure how,” he said. Therefore, “we don’t just give founders capital; we also directly introduce them to potential customers and provide hands-on [go-to-market] guidance,” he added. The fund’s limited partners include Google, along with operators from NVIDIA, Amazon, and Snowflake, totaling over 150 limited partners across a wide range of companies.
Georges said, "Many funds have operator networks. We want to be the truly useful one."
George's investment thesis is based on his assessment of how companies are changing the way they adopt AI. He says, "The experimentation sandbox phase is coming to an end."
“Companies are now highly focused on unit economic efficiency,” he noted. “They’re no longer just paying for open-ended chatbots; they’re willing to pay for deterministic solutions. Real value comes from solutions that are deeply embedded into traditional workflows and actually perform the work”—such as automated code reviews and parsing legal documents.
Georges is preparing for a world where enterprises no longer purchase SaaS tools per user seat. “We will buy completed tasks and outcomes driven by multi-agent workflows,” he says.
To this end, BAG Ventures seeks to invest in companies whose core teams have previously worked together, already have a minimum viable product, have at least one partner, and have a “clear monetization path within 24 hours.”
He also seeks to invest in products that are deeply integrated into enterprise workflows and can access proprietary data that cannot be scraped. As leading AI labs release more products themselves, Georges believes that simply building a technically sound startup product is no longer sufficient in the long term: “If a startup is just a thin wrapper around the API of a frontier model, it will quickly be eliminated.” This is why they look for teams that are deeply embedded in enterprise workflows and control proprietary data that cannot be scraped. “We want companies to have an intent layer and sufficient customer lock-in to survive the next major model release.”
The fund is also focused on startups selling products to highly regulated industries, where data privacy requirements may demand greater specialization. “This means protecting internal data flows, establishing acceptable use guardrails, and deploying continuous automated red team testing,” says Georges. “We’ve seen this approach work well with our portfolio company, Defendremate.”
Georges also stated that businesses will need identity and access management tools designed for non-human employees, such as AI agents. “Startups that can build the next layer of ‘zero trust’ architecture and orchestration channels for agentic systems will fill a massive and highly profitable gap,” he said.