Build custom generative AI experiences for the enterprise with Glean Apps and Actions

One of the most exciting prospects emerging from the rapidly advancing field of AI is the idea of building your own agent—a generative AI system with a specified domain of expertise, and the ability to take action on a user’s behalf. Imagine a true workplace assistant that’s capable of not only answering your work-related questions, but also performing tasks like creating IT tickets, delivering follow-up emails, and debugging code, all with an intimate understanding of all the content, context, and people within your organization.

Today, we’re announcing two exciting new features that’ll help our users achieve just that—Glean Apps and Actions. In this blog, we’ll show you what these features are capable of, and how to get started with building your own customized AI experience for work. 

Glean Apps

Glean Apps are generative AI experiences that require no coding to set up and use. Users can create entirely customized generative AI agents tailored to address specific business needs. Each app operates using a specific segment of company data determined by the user, constraining responses to the most pertinent information. These apps ultimately merge an enterprise assistant’s broad knowledge with agents’ specialized skills focused on particular topics. 

After building an app, it can also be published to the Glean App Library to be discovered by users across your organization. Apps can also be targeted for specific people or departments so that it’s only shown to relevant users.

In this segment, we’ll walk you through how to set one up yourself:

• Step 1
  To start creating a Glean App, provide basic information, including its name and display description. This will help your coworkers understand how and when to use the app to answer their questions.

• Step 2
  Next, add the knowledge sources the App can pull from. These can be individual files, folders, or even all of the data in an entire app. If you don’t know the exact name of a file, that’s okay—Glean can help you find it. All of a Glean App’s results are secure and permissions-enforced, so only the people who have access to the source knowledge will get access to it.‍

• Step 3
  Third, add the app’s custom instructions. This tells the app what it should do, and provides important context for it to successfully execute whatever specific task you put forward. We recommend specifying:
  • What sort of task or tasks the app is expected to perform (“you are a competitive intelligence agent for Acme Corp. Your role is to be a colleague to sales representatives and help them understand how Acme’s products compare to those of its competitors.”)
  • How the app should structure its response (“whenever a user asks a question comparing two products, always respond in tabular form. That table should include a column that lays out why Acme is superior to the competitor.”)
  • What action to take if the app can complete its task (“Always offer to compose an email to a prospective customer that includes this information”)
  • What to do if the app can’t complete its task (“If you cannot answer the question, instruct the user to contact alice@acme.com”)

• Step 4
  If relevant, specify which actions the app can take on a user’s behalf from the company’s action library. Each app can have up to 5 actions.

• Step 5
  Finally, specify which users have access to your app, either by individual or by team, and how it’s triggered. After that, you’re done! The app is ready, and can be published in the Glean App Library for your coworkers to use as well.

Actions

When it comes to a true workplace assistant, we wanted AI to do more than just answer questions. Actions, integrated into Glean Apps (including Glean Assistant), enable Glean to take concrete actions on a user’s behalf. Automate specific parts of your day-to-day workflow, set parameters and expectations, and let Glean handle the busy work for you. 

Here’s a few functions that Actions can currently perform: 

• Read
  Sometimes, the information needed to answer a user’s question exists outside of Glean’s usual index. Using read Actions, Glean Apps and Assistant can get information from sources such as calendars, code repositories, and data warehouses. For example, a user can ask for a code template, and Glean can generate one based on internal documentation from Confluence, chat messages in Slack, and code from GitHub.‍
• Write
  As the name implies, write Actions enable Glean Assistant to add information to company systems. For example, a customer success manager can ask Glean to summarize a conversation from Gong and use that to create a feature request ticket in Jira for the R&D team to review. In another case, a sales rep can use our Gmail action to easily follow up with prospects when they are looking for answers that require a deep dive. ‍
• Build
  Glean also enables users to build custom actions for external applications via the Actions SDK. Take Glean’s framework for building task and workflow automation, and integrate it into the applications you love.

All Glean Actions incorporate the content of files and conversations into their output, ensuring results are as context-rich as possible. Actions also follow Glean’s human-in-the-loop design. This way, users are always aware of what actions are taking place before they happen, and require explicit user approval before any actions are performed. 

AI built just the way you need it

What makes AI so exciting is the sheer scope of what it’s capable of. However, AI agents and assistants require the right foundations and guardrails to ensure great results, along with a way for users to customize them to fit the particular needs of each unique workflow. 

Glean Apps and Glean Actions enable everyone to build domain-specific generative AI experiences with no code or manual setup required. Most importantly, each Glean App and Glean Action operates according to Glean’s robust framework for enterprise AI, ensuring freshness, security, and relevancy for every result. 

Get started today with a truly enterprise-ready AI experience—start building, or get a demo to learn more!