Glean

I ❤️ Glean. I also pity the poor internal champion who has to sell it as “better internal search.” That feels technically true in the same way a dishwasher is “a wetter cabinet.” I use Glean every day at Zapier. It helps me find context, understand history, avoid bothering the same five people, and increasingly, it acts as a context layer for agents I’m building in Zapier’s internal agent harness, which is in closed beta right now. And that is where the ROI conversation gets weird. Because the value is real. But if the pitch becomes “people searched more,” I think we are measuring the receipt instead of the thing we actually bought. Some rough thoughts I’m playing with: 1. Measure the wandering that stopped happening. How often did someone get stuck before finding the right doc, owner, prior decision, or example? 2. Measure the interruptions that disappeared. How many “do you know where this lives?” Slack messages no longer need to happen because the context is findable? 3. Measure agent and automation velocity. How many people started building a useful automation or agent and got blocked because they could not find the right process context, policy, field definition, workflow history, or edge case? Then measure that again after Glean becomes part of the context layer. 4. Measure duplicate work avoided. How often did someone discover that the thing they were about to build, write, document, or decide already existed somewhere? The hard part is that this kind of infrastructure removes a tax most companies were never measuring. Before Glean, nobody had a dashboard for “minutes spent spelunking through Slack like a corporate raccoon.” But that time was never free. I’m genuinely interested in how others are thinking about this, especially as these tools become less “search box” and more “context infrastructure for AI work.” Jonathan Goldberg, you’ve done an amazing Glean rollout at HHX and I’d love your take. How are you measuring the value beyond usage?

How does a major infrastructure firm get back thousands of hours of lost productivity? For McCarthy Holdings, Inc., holding company of McCarthy Building Companies, Inc., it started with giving teams better context, so people could find what they needed faster, spend less time digging, and automate more of the work that used to be manual. Here’s what that looks like with Glean: ⏱️ 2 hours saved per employee, weekly, company-wide 🧰 $2.7M+ in owner change orders reviewed and processed by custom agents 📈 90% active adoption across the organization ➡️ Swipe through for a look at their transformation with Glean https://lnkd.in/g9ZgUJax

Glean has launched independent agents, which are AI coworkers built on its enterprise context layer that act across Slack, Jira & Teams rather than waiting to be summoned, with identity, memory, proactivity, and accountability built in. The On-call Assistant in Slack is a sensible first use case: escalations are bounded and well-instrumented, making the value easy to see. The architectural story is what enterprise buyers will pay attention to. Glean builds agent identity and permissioning around its knowledge graph first, then enables agents operate across surfaces under that identity, which is a different starting point from collaboration-layer-native approaches such as Anthropic's Claude Tag. Both reflect the same conviction that proactive, persistent AI teammates are where enterprise AI is heading. Which approach proves more durable will hinge on permissions and traceability as agents span more systems. Futurum's survey data shows security and privacy remain the top agentic AI concern (24.1% in 1H2026), with loss of human control close behind (16.3%) & a lack of cross-agent governance nearly tripled year-over-year. As agents take on more autonomous identity, that last category is the one to watch. https://lnkd.in/eAcBxCGs

BREAKING: Glean just launched Independent Agents!!!! Picture the 2am escalation. An alert fires. Someone gets paged, half awake, digging through dashboards and old threads trying to work out what broke and who owns it. Thirty minutes gone before the real work even starts. That is the moment Glean just went after. What they launched - A new class of agents that work alongside your teams instead of waiting to be asked. The first one is an OnCall Assistant. When an alert fires, it reads the escalation, pulls the context, chases a few root cause theories at once, drafts a fix, and tags the right person. By the time a human steps in, the first thirty minutes are already done. It had been running inside Glean's own engineering team before they shipped it. That detail matters more than any demo. Why enterprises should care Every leader I talk to wants agents doing real work. Almost none of them will let an agent loose on production, customer data, or internal systems until they can answer three questions. - Can I control what it sees? Yes. Each agent runs with its own identity and scoped access, not a borrowed employee login. - Can I see what it did? Yes. Open any run and every decision and action is right there. - Can I stop it? Yes. One button shuts it down everywhere. That is the shift. Not raw autonomy. Trust you can audit. My take For two years the pitch was look how much our agent can do. The enterprises writing the big checks were never blocked on capability. They were blocked on control. Glean read the room. They built the guardrails first and the autonomy second, and wrapped the whole thing in the context and permissions layer they have spent six years building. And enterprises are clearly buying it. Glean tripled to 300M in ARR in 15 months. That is not hype money. That is production budget. The agents that win inside large companies will not be the flashiest. They will be the ones a CIO can hand real work to and still sleep at night. That is the whole game now. #data #ai #glean #agents #theravitshow

Claude Sonnet 5 is now available in Glean alongside 30+ other leading frontier and open LLMs in our Model Hub. Try it for your work today!

AI is only a true coworker if it can act autonomously and collaborate where work gets done. Our new independent agents have their own identity and context, consistently learn, and can work independently while staying within the permissions, policies, and guardrails your organization sets. That’s what makes them different from other agents. They have their own identity, memory, proactivity, and accountability built in, so they can take action in line with how you and your company actually work. We’ve been using this internally to transform engineering on-call. When an alert fires, the agent investigates, drafts a fix, and tags the right owner, so by the time the on-call engineer steps in, the first 30 minutes of work are already done. Read more: https://lnkd.in/g43b-CKy

90% supervising the AI. 10% actual output. 0% of the time savings you were promised. Sound familiar? That's botsitting, and you might be doing it right now. Digital workers spend 6.4 hours a week botsitting. But where do those hours go? The AI Daily Brief dug into this topic and our Work AI Index report – link to listen in the comments.

Our cutest coworkers clocking in 🐾💼

I cannot wait to present this August at Glean:GO, Glean's annual customer conference in San Francisco. If you are an avid Glean user and want to get up to speed on the latest in enterprise search, context, and AI capabilities, attend Glean:GO!