Glean’s AWARE framework is a practical model for governing AI agents: it focuses on who an agent is acting for, the work context it can access, the guardrails that keep it in scope, real‑time risk checks, and ecosystem observability. Paired with Glean Protect, it helps enterprises deploy agents with clearer controls and oversight.

AI agents are moving fast—from “help me find a document” to orchestrating workflows, pulling data from multiple systems, and taking actions on a user’s behalf. That jump creates a new security problem: traditional app security focuses on access. Agents require governance of behaviour, intent, and risk in motion.

Glean’s answer is the AWARE framework, paired with new and evolving Glean Protect capabilities designed to make enterprise AI safer to deploy and easier to govern.

What is the AWARE framework?

AWARE is a governance framework for the agentic era. It breaks AI agent security into five practical pillars:

• A — Actor intent: Treat agents as first‑class identities. Be explicit about who the agent represents, what role it has, and the scope of actions it’s allowed to take.

• W — Work context: Define which systems, datasets, and workflows the agent can use—based on real business context, not broad access.

• A — Autonomous guardrails: Put boundaries around what an agent can do (and what it must never do), including approvals where needed.

• R — Real‑time risk: Evaluate risk at the moment an agent acts—especially when the situation changes.

• E — Ecosystem observability: Make agent activity visible: logs, traceability, and monitoring across the tools agents touch.

This isn’t theory. It’s a checklist you can use to design agent policies your teams can actually implement. (glean.com)

What does Glean Protect do?

Glean Protect is positioned as the security and governance layer for using AI at work, including agents. In practice, Glean describes Protect as supporting:

• Permissions-aware access: Agents follow existing access controls, so they can only see and act on what the user is already allowed to access.

• Policy controls for agents: You can control who can create agents, how they’re shared, and how they operate within your rules.

• Sensitive data protections: Capabilities such as detecting and auto‑hiding sensitive information to reduce oversharing risk.

• Governance controls: Tools to enforce policies and support compliance requirements as you scale.

These capabilities are aimed at reducing the two most common enterprise risks: oversharing and over‑autonomy. (glean.com)

How AWARE + Protect fit together

Think of AWARE as the “security architecture” and Protect as the set of mechanisms that help you operationalise it.

• AWARE tells you what to define: identity, scope, guardrails, risk checks, observability.

• Protect gives you tools to enforce permissions, reduce sensitive data exposure, and add agent controls.

Together, they support a more mature approach to agent rollout: pilots with governance, then safe expansion.

A practical deployment approach for CIOs and CISOs

If you’re trying to deploy AI agents without creating shadow automation, here’s a straightforward way to start.

Step 1: Pick one workflow that matters

Choose a workflow with clear value and contained risk (e.g., IT helpdesk triage, security ops summarisation, sales enablement). Aim for measurable outcomes.

Step 2: Define “Actor intent” like you would for a service account

• Who does the agent represent?

• What’s the business purpose?

• What actions are allowed, and what actions require approval?

This is where many pilots fail: the agent is treated like a chatbot rather than a governed identity.

Step 3: Lock down work context (least privilege by design)

Define the minimum set of sources the agent needs. If the agent doesn’t need access to a system, it shouldn’t have it.

Step 4: Add guardrails that match the risk

Examples:

• “Draft only, do not send” for customer‑facing comms.

• “Write to a staging area” for HR or legal content.

• Human approval for actions like ticket closures, access requests, or workflow changes.

Step 5: Put observability in place before you scale

You should be able to answer:

• What did the agent access?

• What did it output or change?

• Who triggered it?

• What policies applied?

This is how you move from experimentation to assurance.

Example: governed agent deployment in practice

Glean’s AWARE guidance includes examples that frame agents as identities with scoped access (e.g., an agent provisioned under a caseworker’s identity, limited to specific records and a staging area). This kind of scoping is a useful pattern for any regulated or high‑risk process. (securityjournaluk.com)

Where Generation Digital can help

Most organisations don’t struggle with “getting a demo”. They struggle with building a repeatable, defendable deployment pattern.

Generation Digital can help you:

• Design an agent rollout plan aligned to AWARE

• Define policies and guardrails that security and the business can both support

• Set up governance processes so pilots can scale safely

Related: Explore Glean: /glean/

Summary

AWARE provides a practical structure for governing agent behaviour, not just data access. Glean Protect adds controls designed to keep agents permissions‑aware, reduce sensitive data exposure, and improve oversight as you scale. (glean.com)

Next steps: If you want to deploy AI agents with confidence—without slowing the business—talk to Generation Digital about an enterprise rollout plan: https://www.gend.co/contact

FAQs

1) What is the AWARE framework?
AWARE is Glean’s framework for governing AI agents using five pillars: Actor intent, Work context, Autonomous guardrails, Real‑time risk, and Ecosystem observability. (glean.com)

2) How does Glean Protect enhance security?
Glean Protect focuses on enterprise AI governance controls such as permissions-aware access, policy enforcement, sensitive data protections, and agent guardrails (including controls around who can create and share agents). (glean.com)

3) Who should use AWARE and Glean Protect?
CIOs, CISOs, and platform/security leaders who need to move from AI pilots to scaled deployment with clearer governance, oversight, and compliance alignment. (glean.com)

4) What’s the biggest mistake in early AI agent rollouts?
Treating agents like a chat interface instead of an identity with scoped access, guardrails, and monitoring. AWARE is designed to reduce that gap. (glean.com)

5) What should we do first if we’re starting from zero?
Choose one contained workflow, define agent identity and scope, apply least‑privilege work context, and implement logging/observability before expanding. (glean.com)