Goldman Sachs estimates that "agentic AI could account for 60% of software market value by 2030." That figure, lifted from the briefing material for a KnowBe4 webinar, lays out a simple, immediate fact: organizations are not just adopting tools, they are onboarding fast-moving digital colleagues whose behavior and risk profile differ from human workers.
What is changing: "digital colleagues" and a new perimeter
The webinar framing uses the term "digital colleagues" to describe autonomous AI agents that now act inside corporate environments. Unlike human logins, these agents operate at machine speed and can make the organization's digital perimeter more porous if controls remain human-centric. The material warns that as organizations move from blocking AI to building with it, they introduce a new dimension of security risk that requires revised oversight and interaction security.
Why prompt-engineering can be the new social engineering
KnowBe4's session, led by Martin Kraemer, KnowBe4 CISO Advisor, emphasizes that "AI agents can be prompt-engineered just as easily as humans can be socially engineered." The difference is procedural: autonomous agents do not come with an inherent grasp of a company’s risk tolerance. That mismatch demands "a new approach to oversight and interaction security," in the words of the webinar description, rather than relying solely on traditional behavior-focused strategies aimed at human users.
Recent AI threats and how they evade traditional controls
The webinar promises a "look at recent AI threats, how they bypass traditional controls, and how they could have been avoided." While the briefing does not supply case-level details, it explicitly connects the rise of AI-based attacks to the growing inadequacy of legacy controls. The implication put forward is straightforward: existing defenses tuned to human patterns can be blind to attack vectors that exploit prompt design, automation chains, or unexpected agent behavior.
Practical steps: reduce Shadow AI and enable sanctioned use
Attendees are told they will learn "practical steps you can take immediately to secure AI adoption across your organization." Central to those practicalities is managing the interaction between employees and AI to eliminate what the brief calls "Shadow AI" while still encouraging sanctioned, governed use. The goal framed by the webinar is not to forbid AI, but to shift organizations from reactive blocking toward deliberate, secure building.
Agents supervising agents: speed outpacing human grasp
Another topic flagged for discussion is the risk posed when agents supervise other agents "in production, where the pace of interaction exceeds human ability to grasp." The webinar material highlights the need for specific strategies to manage multi-agent systems in live environments, recognizing that layered automation can generate risks that human operators cannot track unaided.
What this means for security teams, procurement leaders, and employees
- Security teams: Must adopt oversight models and interaction security that assume agents will act at machine speed and be vulnerable to prompt manipulation.
- Procurement and architects: Need to prioritize sanctioned, governable AI deployments to prevent Shadow AI and ensure controls travel with the technology into production.
- Employees and line managers: Should be brought into governed workflows so that human–AI interactions are predictable, auditable, and aligned with organizational risk tolerance.
The KnowBe4 webinar, presented by Martin Kraemer, packages these issues into actionable content: why prompt engineering matters, where recent AI threats have succeeded, immediate defensive steps, how to stamp out Shadow AI, and how to govern multi-agent systems. Attendees are promised that they "will leave with clear next steps for navigating the convergence of human and AI risk," a concise encapsulation of the brief's core promise.
https://www.govinfosecurity.com/webinars/your-organization-ready-to-scale-ai-securely-w-7064
