The enterprise AI agent market is fragmenting into three distinct approaches: learning from real customer interactions, building independent infrastructure for agent development, and establishing identity standards for autonomous agents operating across the open internet. Rather than relying solely on frontier AI models from companies like OpenAI and Anthropic, organizations are increasingly investing in specialized platforms that give them control over how their agents learn, operate, and identify themselves (Sources 1, 2, 3).

What's Driving Enterprises Away From Closed AI Models?

Companies are growing wary of depending on proprietary AI systems for mission-critical tasks. The risks are real: data privacy concerns, sudden service discontinuations, and the fear that AI labs might eventually compete with their customers. This shift is creating an opening for startups that offer alternatives.

Encore AI, which raised $30 million in Series A funding, exemplifies this trend by focusing on a completely different angle. Rather than building agents from scratch, the company analyzes recordings of successful customer interactions between employees and clients, then trains AI agents to replicate those winning approaches. The platform collects call recordings, emails, and text messages, connecting them to customer relationship management (CRM) systems to identify which conversation tactics actually moved deals forward.

"The agent we build is a package of many different playbooks that have worked throughout the process. Sometimes our agents even tell the jokes that the relationship managers are telling, or give the anecdotes or examples that the relationship managers are giving, because we literally run by the playbooks that we see working," said Dvir Ginzburg, CEO of Encore AI.

Dvir Ginzburg, CEO at Encore AI

Encore's approach, which Ginzburg calls "interaction mining," has resonated with financial institutions. The company now serves more than 40 enterprise customers globally, with annual recurring revenue increasing more than 5 times since its seed round less than 18 months ago.

How Are Enterprises Building Their Own AI Agent Infrastructure?

Meanwhile, Prime Intellect is taking a different approach by providing the underlying tools and computing power that allow enterprises to train their own agents without relying on external AI labs. The startup raised $130 million in Series A funding at a $1 billion valuation, backed by investors including Radical Ventures, Nvidia Ventures, and Intel Capital.

Prime Intellect's platform functions as a full-stack solution for AI agent development. Rather than forcing customers into an all-or-nothing arrangement, the company operates like a marketplace where organizations can pick and choose specific tools they need. The platform includes compute access, a reinforcement learning framework (a technique that iteratively rewards successful task completion and penalizes errors), and evaluation tools.

• Compute Access: Organizations gain dedicated computing resources to train and deploy their own models without relying on external AI providers.
• Reinforcement Learning Framework: Companies can refine AI models for specific business tasks by rewarding successful outcomes and learning from failures.
• Evaluation Tools: Built-in systems help enterprises measure whether their agents are performing as intended before deployment.

The results speak for themselves. Ramp, a fintech company, used Prime Intellect to build an agent that searches spreadsheets for answers. The result outperformed frontier AI models on accuracy while running faster and costing a fraction of the price. Prime Intellect has already reached an annualized revenue run rate of $100 million, with customers including Zapier and Flapping Airplanes.

"It shouldn't just be a few nerds in a glass tower in San Francisco that have the capability to train AI models. It should be every enterprise, every nation state," said Vincent Weisser, co-founder and CEO of Prime Intellect.

Vincent Weisser, Co-founder and CEO at Prime Intellect

Who's Building the Identity Layer for Autonomous Agents?

A third critical piece of the puzzle is emerging: how do you identify and hold accountable AI agents operating autonomously across the open internet? Vint Cerf, one of the architects of the internet itself, is now advising Innovation Labs on exactly this problem.

Most AI agents today operate within closed, proprietary systems. But as businesses envision a future where agents interact directly with other agents across the internet, a fundamental question arises: how do you know who or what you're dealing with? Innovation Labs has proposed DNSid, a system that creates identities for agents and links each one to an existing internet domain name using cryptographic proofs to log its registration over time.

"I felt like I might be able to help them in a period of time when naming and identification is becoming increasingly important. This is largely triggered by the notion of AI agents and the question of what authorities they have, where they have derived those authorities, who is accountable for the behavior of an agent in this context, and where and how its identity is established, and why you'd trust it," said Vint Cerf.

Vint Cerf, Advisor at Innovation Labs

The stakes are high. Without clear identity standards, agents from different companies won't be able to interoperate with each other, fragmenting the potential agentic economy. Cerf notes that the key to wide adoption will be functionality, just as it was with TCP/IP, the foundational protocol of the internet.

What Are the Key Takeaways for Enterprise Leaders?

The convergence of these three trends reveals a fundamental shift in how enterprises will approach AI agents. Rather than treating agents as black boxes managed by external providers, organizations are investing in platforms that give them control over agent training, deployment, and accountability. Encore AI's focus on learning from real customer interactions, Prime Intellect's infrastructure for independent model training, and Innovation Labs' identity standards represent three complementary solutions to the same underlying problem: how do enterprises build trustworthy, autonomous AI agents that serve their specific needs without surrendering control to external providers (Sources 1, 2, 3).

For financial institutions, customer service teams, and enterprises handling sensitive data, these alternatives offer a path forward that balances the power of AI agents with the need for transparency, control, and accountability.