Volume 343, Issue 3 · 7 articles from 4 issues
Glean explains Jev, a model for typed decisions, and benchmarks it on classification, routing, reranking and citation judging.
Glean's engineers describe Jev, a 'System One' model that takes context plus a fixed set of questions and options and returns choices, scores and probabilities instead of generated text.
The post compares Jev with alternative approaches for producing labels, scores and yes/no answers, and acknowledges skeptics' points about classifiers, structured outputs and small models.
Glean benchmarks Jev across four enterprise uses: classification, routing, reranking and citation-support judging.
Glean explains how it scans, sandboxes and governs AI agent skills to block prompt injection, data exfiltration and over-permissioned access.
Glean describes how it secures skills, the packaged tools and instructions that extend AI agents and could otherwise hide over-permissioned access or malicious instructions.
Its scanner evaluates the whole package for prompt injection, data exfiltration risk and excessive agency at creation, upload, import, sharing and publish time, rescanning on every change and issuing a verdict of pass, review recommended or blocked.
At runtime, skill code runs in an isolated sandbox, and a credential broker mediates access to external systems so credentials never sit inside the skill.
Canva shared at Glean:GO 2026 how it moved through AI everywhere, AI impact and AI native stages with people at the center.
At Glean:GO 2026, Canva's AI and automation leads described a three-stage journey of AI everywhere, AI impact and AI native, with a new emphasis on AI accountability and a goal of using AI to empower people rather than replace them.
The internal journey began in October 2023 with Glean inside Canva's Canvaworld directory.
As employees experimented with other tools, Canva made Gemini, ChatGPT and Claude available by request alongside Glean, moving from approval to company-wide announcement in six days.
At Glean:GO 2026, Expedia Group explained how it chose and rolled out an enterprise AI front door with IT guardrails.
At Glean:GO 2026, Expedia Group's employee digital experience and AI enablement leads described how they gave employees one clear place to start with AI, after embedding AI into existing Microsoft, Slack and Zoom contracts still left employees without connectivity across enterprise context.
A formal RFP evaluated AI platforms on security and governance for IT and on employee trust and adoption, and Glean was selected.
A cross-business Pioneers Group stress-tested tools early, and success was measured by behavior change rather than usage numbers.
Glean's skill reflection lets reusable AI skills learn from usage and surface governed improvements over time.
Glean says skills package reusable capabilities that teach AI to perform tasks such as triaging tickets, drafting escalations or reviewing contracts, but they are frozen snapshots that go stale as processes change.
Skill reflection lets skills learn and improve from AI interactions in a governed way, watching during use for reusable signals such as a repeatedly missing step, a correction that maps to an instruction, or patterns in how users steer the work.
At the end of a session it evaluates that evidence, aggregates learnings and surfaces suggested changes.
Glean launches Glean for Data Analysis to help employees answer why metrics moved using data and company context.
Glean has introduced Glean for Data Analysis, aimed at the follow-up questions that dashboards cannot answer, such as why a metric moved.
Glean says the answers are often scattered across systems like Salesforce, Slack and support tickets, or held in people's heads.
General-purpose AI assistants may not know company-specific data definitions.
Glean for Data Analysis combines data insights with enterprise context so every employee can investigate questions.
Glean introduces headless agent building so developers can build, test and publish Glean agents from coding tools.
Glean has introduced headless agent building, which lets developers build, test and publish Glean agents from coding environments such as Claude Code, Cursor and Codex.
Glean says agents doing consequential work need software rigor such as version control, testing, review and rollback.
Building agents as code gives developers that control and more deterministic workflows.
Glean adds the enterprise permissions, governance and auditability these agents need.