Security

CISOs Must Build Trust and Assurance Teams for AI Governance

Forrester research shows security leaders need new organizational structures as agentic AI outpaces traditional cybersecurity models.

Omega Editorial · November 19, 2026 · 4 min read

Key takeaways

  • 37% of enterprise AI projects fail before production due to security, privacy, and trust concerns, yet economic pressure continues driving rapid agent deployment.
  • CISOs must produce verifiable evidence that AI agents operate within policy boundaries, even when deployed by business units outside security's direct control.
  • Forrester's trust and assurance organization (TAO) model distinguishes between assurance (what security produces) and trust (what stakeholders decide).
  • Security leaders face both authority gaps (responsibility without control) and capability gaps (authority without evidence-producing systems).
  • AI governance roles are being redistributed across enterprises now—security leaders who wait will have their responsibilities defined for them.

CISOs Face Organizational Crisis as AI Agents Proliferate

Chief Information Security Officers are confronting a structural problem: their organizations weren't designed for the world they now operate in. As enterprises deploy AI agents at scale, security teams must not only protect systems and data but also demonstrate that automated decision-making stays within approved boundaries. That dual mandate requires fundamentally rethinking how security functions are organized, staffed, and measured.

Why it matters

The gap between AI deployment speed and governance capability creates genuine business risk. When 37% of enterprise AI projects fail before production due to security and trust concerns, yet economic pressure continues driving adoption, security leaders need authority and capabilities that match their expanded responsibility. Organizations that fail to evolve their security function risk having AI governance responsibilities distributed across the enterprise by default rather than design.

The Old Security Model Is Breaking

According to research from Forrester, business units are deploying agents faster than they can govern them. The firm's State of AI Survey found that data privacy, security concerns, and trust in agentic systems rank among the greatest barriers to AI adoption. Yet boards continue pursuing automation because the economic incentives favor productivity and scale.

This creates a challenge security teams haven't faced before. CISOs must produce defensible evidence that agents, models, and automated decisions operate within policy—even when the business units deploying them sit outside security's direct control.

Meanwhile, security's own work is being automated. As agents handle detection, triage, and response tasks, teams must determine what those systems can do, when human intervention is required, and how to validate results. Richard Watson, global cybersecurity consulting leader at EY, noted that for CISO organizations, "the onus is on them to assure that the work is automated rather than doing it themselves."

Introducing the Trust and Assurance Organization

Forrester proposes a new model: the trust and assurance organization, or TAO. This represents a CISO-led evolution of the cybersecurity function designed for an era where AI-driven systems make and automate business decisions.

The TAO uses verifiable controls, governance, evidence, and transparent reporting to attest to the integrity, security, privacy, and responsible use of data and AI. Critically, Forrester distinguishes between assurance and trust. Assurance is what security produces: technically enforced controls, continuous validation, and examinable evidence. Trust is what stakeholders decide after examining that evidence. Security can earn trust but cannot certify it.

Testing for Capability Gaps

Forrester recommends CISOs trace one production AI use case and ask three questions: Who owns the business outcome and accepts residual risk? Who has authority to change the agent's permissions and boundaries? What evidence shows those boundaries hold?

This exercise typically exposes two types of gaps. An authority gap exists where security must explain outcomes but has no control over the decisions that produced them. Closing this requires executive-level discussions about risk ownership and reporting lines. A capability gap exists where security has authority but cannot produce evidence because functions still depend on periodic reviews and manual processes.

Building TAO Before It's Too Late

Enterprises are already redistributing responsibility for AI governance, identity, data protection, and customer trust. Roles left undefined get assigned through momentum rather than intentional design.

Forrester's research, first reported by analyst Madelein Fombad and colleagues, provides a reference organizational chart, explains how eight security functions must change, and replaces activity-based metrics with measures that show whether controls actually work.

The message for security leaders is clear: evolve the organization now, or watch AI governance responsibilities get distributed across the enterprise without your input.

Details in this article were first reported by Forrester Research.

  • #ai governance
  • #ciso
  • #cybersecurity organization
  • #agentic ai
  • #trust and assurance
  • #forrester

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in Security