Quick Facts
- Only 3% of companies are successfully scaling agentic AI across multiple departments, even as 62% actively experiment with it.
- 88% of organizations confirmed or suspected security incidents tied to AI agents, and only 14.4% have full security approval for their deployments.
- BAND exited stealth with $17 million in seed funding to build coordination infrastructure for multi-agent AI systems.
Enterprise AI agents are everywhere. The infrastructure to govern them is not. That gap was front and center at VB Transform 2026, held July 14-15 in Menlo Park, where five startups presented solutions to the core failures of enterprise agent deployments: agents that cannot communicate, cannot be trusted with permissions, and cannot be audited when things go wrong.
The numbers behind the problem are stark. IDC data from November 2025 shows 50% of organizations already had 10 or more agents in production. Yet only 3% of companies are successfully scaling agentic AI across multiple departments. Gartner projects that 40% of enterprise applications will embed task-specific AI agents by end of 2026, up from fewer than 5% in 2025. By 2029, IDC forecasts the number of actively deployed AI agents worldwide will surpass 1 billion.
Adoption has outpaced governance by a wide margin. A Dataiku and Harris Poll survey of 800 global data leaders found 86% of organizations rely on AI agents in daily operations. Most lack the orchestration and governance infrastructure to manage them at scale.
The Security Gap
The security picture is worse. A Gravitee survey of more than 900 executives found 88% of organizations confirmed or suspected security incidents tied to AI agents. Only 14.4% of teams have full security approval for their deployments. Just 22% treat agents as independent identities. The rest rely on shared API keys.
That credential-sharing habit carries a measurable cost. Organizations that allow credential sharing anywhere experienced a security incident or near-miss at a 63.5% rate. Companies where every agent has its own scoped identity saw that figure drop to 40.9%. The 2026 CISO AI Risk Report found only 16% of organizations effectively govern AI access to core business systems. Among companies with revenue above $1 billion, 64% reported losses exceeding $1 million tied to AI system failures in 2025.
Gartner also flagged what it calls “agentwashing” — the widespread mislabeling of AI assistants as agents. VB Transform 2026 data put the scale of that problem in concrete terms: 71% of enterprise agents are actually chatbots, 27% have no kill switch, and 70% use service-account access.
Startups Moving In
BAND, also known as Thenvoi AI Ltd., is building a coordination layer for multi-agent systems. The core problem the company addresses is simple: an agent built on LangChain cannot easily hand off tasks to one built on CrewAI. A Salesforce-embedded agent has no native way to coordinate with a custom Python script running on a private cloud. Existing platforms like Slack or Discord were built for humans, not for non-deterministic software systems.
“In order for agents to become real players in the global economy, they need ways to communicate, just like humans do,” said co-founder and CEO Arick Goomanovsky. “The communication solutions we have today for systems don’t work for agents, because agents are non-deterministic creatures. It’s not just about API integrations.”
BAND exited stealth with $17 million in seed funding backed by Sierra Ventures, Hetz Ventures, and Team8. The platform supports autonomous workflows running 8 to 20 hours and is compatible with A2A and MCP protocols. Humans can join agent conversations in real time. Goomanovsky previously co-founded Sygnia, sold to Temasek for $250 million, and Ermetic, acquired by Tenable for approximately $300 million.
Conifers is targeting the security operations problem directly. Its CognitiveSOC platform connects threat intelligence, threat hunting, detection engineering, investigation, and remediation into a single operating layer. “The biggest challenge defenders face today is that they’re still running at human speed, but adversaries are running at machine speed,” said CEO Tom Findling. Malicious campaigns that once took months now take hours.
The business case for all five startups rests on the same underlying reality. Agent deployment is already happening at scale. The tooling to make those deployments secure, auditable, and coordinated has lagged behind. Companies building that infrastructure are stepping into a market where the demand is already proven and the failures are already documented.
This article was written by an AI agent. Spotted an error? Send a correction and we will fix it.