Using Glean Chat as a software engineer – New code and conversations

Garvit Juniwal

Engineering

Krishnan Raghavan

Engineering

As Software Engineers at Glean working on building assistant features like Gleanbot and the platform services they rely on, there’s plenty to do and often not enough time to get it all done. Some of the busiest work on our plates involves two things – getting started with new code, along with gathering the company-specific context required to do so. 

Writing code is a highly collaborative and iterative process. It’s incredibly helpful to have Glean Assistant, armed with all the right content and context, provide us with proactive responses to our questions – or even take the first few passes on generating code. 

Writing code with company context

Let’s say that we’re trying to implement a feature that requires copying a row in a table to a new key. If we ask ChatGPT-4, we can expect a reasonable answer, but it’ll probably be too generic for us to actually use in a work context. It doesn’t know which database we use, or the schema of the table that’s being referenced.

ChatGPT provides a good response, but not good enough for work

However, when we ask Glean Assistant the same question, it can use all its indexed company knowledge. By reading PRs and code files from our repo, Glean knows which database we use at Glean (MySQL) and the KeyVal table schema. It generates the exact SQL query we can plug into the backend code.

Glean provides a context-rich response that gets me back to work quickly

Being able to generate immediately applicable code that’s augmented with company-specific knowledge and requirements enables us to move quickly and securely! 

Proactive responses make for magical moments

Another excellent use of Glean Assistant is its capability to understand conversations and context within Slack. Let’s say that I was curious about the status of an internal cloud project. I had no idea who to ask, so I posted a question in a public Slack channel to see if anyone would bite. Given that we have a global team, I expected it to take at least a few hours for the right person to respond.Lo and behold, Glean Assistant came to my rescue almost immediately!

Glean, unprompted, delivers me a comprehensive and correct answer

Glean is present in Slack and continuously monitors public channels for questions it can answer. In this case, I got an authoritative and correct response instantaneously without remembering to ask Glean explicitly. The answer was presented in a PR description about bugs we faced when setting up the project, which shows how Glean can bridge sources of knowledge (Slack and GitHub) to provide a comprehensive answer enriched by context sourced from across the entire company. 

Relying on Glean Assistant explicitly

Supporting all the complex components of the software we build is a core part of the job; much of that happens in Slack. At Glean, we have #help-<component> channels, where component owners monitor for questions and respond appropriately. 

In many cases, questions are often repeated or require gathering additional information through following some debugging steps. Instead of typing the response out over and over again, we often have Glean Assistant assume the responsibility of responding. Since Glean has indexed knowledge from across all apps, it’s capable of responding to multi-person conversations in whichever thread it was asked to respond in. Generated answers will have citations and links to any additional resources and source material.

Glean consolidates multiple sources and conversations to deliver a comprehensive solution

It’s like having a teammate in Slack that has memorized all company knowledge – and responds instantaneously!

Generative AI that’s ready for the enterprise

Glean Chat has proven to be an incredible asset for enhancing productivity. It has simplified and expedited coding tasks, merging the functionalities of a code assistant and a workplace aide to boost my efficiency and output significantly. Its ability to comprehend and craft code by leveraging context and existing documentation makes it an essential tool for any software engineer in search of a genuine workplace companion that excels in both corporate and programming knowledge.

{{richtext-cta-component}}

If you’re looking to supercharge your or your team’s productivity with a generative AI solution that’s truly enterprise-ready, sign up for a demo today!

Related articles

No items found.