News
White Paper
The Future of Leadership Development: AI-Driven Manager Enablement
In Brief
- Glean introduced a structured, seven stage Agent Development Lifecycle for enterprises.
- New platform features focus on agent governance, context and measurement.
- Today's announcement establishes a new AI battleground: agent lifecycle governance.
Glean today introduced its enterprise Agent Development Lifecycle (ADLC) framework, designed to tame agent sprawl. The ADLC aims to give CIOs a repeatable path to deploy, govern and measure AI agents.
Alongside the framework, Glean released new capabilities across its platform spanning agent building, governance and monitoring. Some features are generally available, while others remain in beta or are coming soon.
ADLC, Updated for the AI Agent World
The Application Development Lifecycle has long given teams a repeatable process for building applications from planning through maintenance. Glean is applying that same logic to a newer challenge: getting AI agents to work reliably inside large organizations.
Worth noting before jumping into the details: the framework assumes teams are starting fresh. For organizations already mid-sprawl, it's unclear where the framework fits. During the same event, one attendee asked the question: should companies audit their existing agents first, or pick a single high-value use case and rebuild it properly? Glean's framework doesn't explicitly address retroactive adoption — a notable gap given that agent sprawl is the very problem it claims to solve.
The framework has seven stages:
Opportunity identifies the business problem before anyone builds anything. Not every task requires an agent, and this stage forces the question early: where will an agent save time or improve outcomes compared with a simpler solution?
Next comes Design, where teams map what an agent should do, which systems it needs to connect to and how it should interact with people.
The company made a deliberate choice to include Performance before development begins. Setting success metrics upfront, rather than after launch, helps organizations avoid the trap of building something technically impressive that no one can prove is useful.
Input gives the agent access to the right company data and context, with the appropriate permissions. An agent that cannot reach the information it needs — or one that can access information it should not — is either ineffective or a liability.
The build and test stage comes next. During Develop, teams use Glean’s platform tools — including the Auto Mode builder, debug views and sub-agent architecture — to implement the workflow. They trace the agent's decision-making step by step to catch problems before they reach users.
Launch involves setting access policies, applying organizational guardrails and distributing the agent through a controlled library. The goal is to avoid the scenario Glean warns about: ungoverned agents proliferating across teams without oversight.
Finally, Monitor & Improve treats the agent as an ongoing task rather than a finished product. Teams track adoption, gather feedback, measure hours saved and refine the agent over time. The phase resembles the maintenance stage of traditional software development, but with a stronger emphasis on business value: did the agent deliver on what the Performance stage promised?
In practice, this final stage may be harder than it sounds. During Glean's live event, a practitioner noted that agents often struggle with the improvement part of the loop — when given feedback, they tend to drift off-task or lose sight of broader context rather than course-correcting cleanly. Glean's upcoming Insights Dashboard tracks performance trends, but tracking and closing the feedback loop are different problems. This is an open challenge across the industry, not just for Glean.
Updates to the Glean Platform
Glean announced a number of new features to support the different stages of the agent lifecycle. It broke them down into three practical needs.
Building Agents Faster
Auto Mode lets users describe what they want an agent to do in plain language instead of having to configure them step by step. New Debug and Trace views expose how agents arrive at decisions, giving builders clearer visibility into failures, reasoning paths and execution flow.
Glean also introduced sub-agents, allowing parent agents to delegate specialized tasks to smaller, focused components with the intention of keeping increasingly complex workflows manageable. Additional updates expand the agent sandbox with secure file handling and code execution, while new event-based triggers allow agents to respond automatically to content updates, scheduled actions and other workflow events.
Controlling How Agents Are Used
The new Agent Library Controls add verification badges, featured agents and departmental categories so organizations can manage which agents are visible and trusted.
Agent Access Policies is still in beta. With it, admins can set organization-wide rules, blocking sensitive content from being processed or restricting which teams can write to critical systems.
Measuring Whether Agents Deliver Value
An updated Agent Insights Dashboard, coming soon, can track adoption, top use cases, estimated hours saved and feedback trends over time.
AI Agent Lifecycles: Build, Govern, Scale
Enterprises are replacing standalone AI features with structured lifecycle systems to build, govern and scale AI agents across workflows.
As recent Reworked coverage noted, agents are shifting from isolated capabilities to coordinated systems for orchestration, security and workflow management, making governance and lifecycle management foundational requirements.
Conference
Jun
2
Gartner Application Innovation & Business Solutions Summit Las Vegas 2026
Register
Webinar
On demand
Retention Is a Communication Problem: New Data from 6,000+ Employees
This session gives leaders a clear diagnosis and practical path to treat retention as an infrastructure problem.
Watch Now
Webinar
On demand
Do Learning Programs Really Work? How to Turn Education Into Engagement In Healthcare
See how leaders are using learning programs to build trust with healthcare professionals and create measurable engagement.
Watch Now
Webinar
On demand
Fuel Your Front Line: Workforce Trends Shaping 2026
Get expert insight into the tools and tactics top frontline teams will need in 2026.
Watch Now
Webinar
On demand
Fix the Disconnect Between Projects and Profit
Your margins are hiding in disconnected data. Let's find them.
Watch Now
Webinar
On demand
AI in Customer Service: Faster Resolutions, Happier Customers
Don’t let rising demand burn out your team. See how to build a smarter, more resilient support org.
Watch Now
Google's Four-Pillar Framework
Google's Gemini Enterprise Agent Platform offers one of the clearest lifecycle models. It breaks it down into four pillars: Build (low-code to full-code tools), Scale (infrastructure for long-running agents), Govern (centralized identity and access control) and Optimize (testing and automated refinement).
Security & Oversight Models
While vendors increasingly agree on the need for lifecycle governance, they differ sharply in how control should work.
Laserfiche's AI Agents inherit the permissions of the initiating user. OpenAI takes more of a marketplace approach with Frontier, assigning each agent an employee ID with explicit access controls. Wrike embeds its AI Agents directly into collaborative work environments, treating them less as standalone systems and more as operational teammates integrated into day-to-day workflows.
What's less clear is what happens when agents span multiple platforms with incompatible rules. Glean emphasizes open APIs, MCP interoperability and support for external models, but its approach still centers on a shared governance and context layer controlled through the Glean platform. Will enterprise AI converge around interoperable standards or will it continue to fragment across competing governance models?
Have a tip to share with our editorial team? Drop us a line: