Investors Ramp up Bets on the Agent Economy

Startup Spotlight

Sphinx Labs is building AI agents designed to replace the human bottleneck inside financial compliance operations. The idea emerged from a simple observation. As Co-founder Alexandre Berkovic explained to PYMNTS, no matter how much compliance software improved, “it didn’t really change the time to review cases, and the reason was that the bottleneck was humans.” Analysts still had to log in, review data and make decisions manually. “The efficiency gains of making a better platform and better data were marginal.”

Sphinx’s approach is to deploy agents that operate directly inside existing compliance systems.

“Sphinx works just like as a human. So you’re going to give Sphinx a username and a password, access to different systems,” Berkovic said. The system ingests company-specific “procedures, policies, SOPs, risk matrices, anything really specific to that company” and executes reviews in line with those internal standards.

Initially, the agents suggest decisions for human approval. “Analysts are going to approve or reject our decision,” he said. Over time, as confidence builds, “we’re going to start automating the decision end to end.”

The company works with “big banks and payments processors in crypto and FinTechs,” particularly in high-volume environments where compliance creates heavy operational overhead. Berkovic said that in some deployments, Sphinx has “been able to automate over 92% of compliance operations,” with the remaining 8% routed to manual review. It has also “been able to find up to 17% more true positives,” detecting suspicious activity human teams might miss.

Sphinx recently raised $7.1 million in a seed round led by Cherry Ventures, with participation from Y Combinator, Rebel Fund, Deel Ventures and Singularity Capital.

Other funding rounds from last week include:

AI Infrastructure and Compute

MatX raised $500 million to build next-generation AI chips designed to compete with Nvidia in high-performance AI workloads, according to Bloomberg.

The company is developing silicon optimized for large-scale model training and inference, positioning itself as part of a growing wave of startups attempting to challenge the dominance of incumbent GPU providers. The round underscores sustained investor appetite for foundational AI infrastructure as demand for compute continues to rise.

On the data layer, Encord secured $60 million in a Series C round to scale its AI-native data infrastructure platform. The company provides tools to manage, curate and annotate multimodal datasets, including video, imagery and sensor data, as enterprises expand beyond text-based AI applications into physical and real-world deployments. As model performance increasingly hinges on high-quality training data, startups like Encord are targeting a core bottleneck in the AI stack.

Enterprise AI and Workflow Automation

Basis raised $100 million in a Series B round led by Accel and Google Ventures at a reported $1.15 billion valuation. The company builds AI agents for accounting firms, automating complex workflows across tax, audit and bookkeeping.

Koah, meanwhile, closed a $20.5 million Series A to scale AI-native monetization infrastructure. The startup is building contextual advertising systems designed to integrate directly within generative AI chat experiences, aiming to create revenue models tailored to conversational interfaces rather than traditional display formats.

Outside AI infrastructure, Honest Health raised $140 million to expand its value-based care platform for health systems. The company partners with providers to improve care coordination, analytics and cost management as hospitals transition away from fee-for-service models. While not exclusively an AI company, its technology-enabled approach reflects continued investor interest in data-driven healthcare transformation.