Business contact management app
AI-powered slackbot
Our annual hackathon saw ideas ranging from healthcare, and fitness all the way to workforce optimization. There was one team of AI experts, however, that wanted to show what they’re capable of.
SHORT
summary
Industry
AI
Location
Croatia
Timeline
1 day
Services
AI development
Team
3 Software engineers
1 Presentation specialist
The goal was to create an AI-powered Slackbot to help employees reach their own version of AI that knows all useful information about the company. You wouldn’t have to read the handbook to understand how the coffee machine works. You could just talk to the bot.
The challenge
It was important to find the right AI language model and to feed all of our info to the bot.
Among the primary challenges the team faced was keeping each conversation private and continuous. At first, the bot didn’t recognize the differences between users and it would forget each conversation.
That’s nothing a good database couldn’t fix.
The solution
We decided the Meta-developed Llama is the best language model for our project. So we fed the model with our entire company handbook.
We’ve also used a database deployed on Heroku to keep giving each individual’s context to the Slackbot every time a conversation starts. That way, we no longer had to worry about mixing contexts and the bot forgetting previous discussions with users.
How we got there
We’ve spent an entire coffee-fueled afternoon creating the integration between Slack, the language model, and Chat GPT. A few hiccups here and there didn’t dissuade us from our goal to win the hackathon and shove it in everyone else’s faces.
As you’ll find out later, not everything went according to our plan.
Our tools
Python
All of our programming efforts were done with good ol’ Python.
ChatGPT
Generative AI doesn’t get much better than ChatGPT 4, which we used to create user-readable answers.
Heroku
To keep the context of previous conversations, we deployed a database through Heroku which the tool would use to pull relevant data every time a user asks a question.
LLaMA
To create tokens that will communicate with GPT, we’ve used LLaMA developed by Meta.
Rome wasn’t built in a day…
but our chatbot was.
When all was said and done, here’s what each employee could do with the Slackbot:
- Ask it any of the usual stuff you ask ChatGPT
- Ask it useful, company-specific information
- Keep the conversation going over days, months, or even years.
Sounds useful right? Sounds like a project that would win the hackathon. Well, no.
A bathroom availability app. That was the winner. An app that uses sensors to tell you when the office bathrooms are available.
A bathroom. Availability. App.
The moral of the story is: The closer your solutions are to your users’ basic needs, the more effective it’s going to be. We guess?
WANT A PRODUCT LIKE THIS ONE? — LET’S TALK
Just fill in the short form below, or skip to the bottom to arrange a chat directly with our client strategy specialist. Either way, we’ll be back in touch as soon as we can to arrange an informal call.
Don’t worry, any information you give us will only be stored for business purposes. Find out more in our Privacy Policy.
Arrange a chat with Ivor, our client strategy specialist.