Case Study
A.R.E.S.
Contents
The Client
The A.R.E.S. product was built for our own team use.
The Challenge
Bring the power of LLM’s to bear in an environment where our team spends a substantive portion of our time communicating: Slack.
Like many companies today, Slack is a go-to platform across the Apollo 21 team. We utilize the environment for communication, document sharing, etc. — so much so that it has become ubiquitous in our (and many other companies’) day to day activities. It’s the one app that our team members are likely to have always-on.
With this in mind, Slack is the ideal place for our team to interact with an LLM without having to break out of our standard application set by visiting the ChatGPT website or mobile application. With this in mind, we created A.R.E.S. — the Apollo Responsive Exploration System — as a Slack application that’s readily available.
The Plan
As LLM’s have become ubiquitous with the introduction of ChatGPT, and the capabilities of such technologies have continued to grow, the Apollo 21 team has begun integrating use of LLM’s for a variety of tasks into our workflow. We use LLM’s for anything from answering questions, finding new services and technologies to explore, and writing first drafts of a variety of communications across our team. As our usage has grown, so has our desire for an easier place to interact (compared to the ChatGPT website and app).
The Solution
Utilize the ChatGPT LLM to create a Slack-based interface for prompting ChatGPT.
A.R.E.S. is a Slack app that brings the power of ChatGPT 3.5 (A.R.E.S. was released before the introduction of GPT 4.0) directly into the Slack environment. A.R.E.S. can be queried via the app-specific channel or by @ mentioning A.R.E.S. in any other channel. In addition to the standard ChatGPT Q&A functionality, A.R.E.S. has the ability to follow on to specific requests when further queried by the user in a Slack thread. This allows us to create continuous, topical conversations and queries (much as one might within the ChatGPT website).
In addition to the previously described functionality, we also connected A.R.E.S. to an early questioning bot to see where we might find limitations in its ability to answer questions. Every day, our WHEN app asks A.R.E.S. a question or two in an attempt to stump the LLM.
To date, WHEN has not successfully come up with an unanswerable question.
The Outcomes
Easier access to a tool that is becoming highly ingrained in the Apollo 21 workflow.
In Their Own Words
Work With Us
Sound like a process that works for you?
Request a Free Brainstorming Session