Success story

Developing a local AI chatbot for the modern workplace

  • AdTech
  • AI transformation
  • eCommerce
  • Healthcare

Administrative tasks can be exhaustive and steal energy and resources from the crew. It’s important to focus on what’s important and save the capacity.

So, we developed a secure AI chatbot tailored to free up staff’s time for their main job. The goal was to create a handy solution that: 

  • provides accurate and real-time responses to HR and security-related inquiries, and  
  • improves the staff productivity to make their routine effortless and more efficient.

Project numbers

15
queries/sec

can be handled by the new chatbot

1
month

took us to develop and deploy the MVP

3-7
seconds

to get an accurate response

Business challenges

Time-consuming administrative and HR processes, impacting overall productivity. 

Lack of staff engagement due to cumbersome access to internal data. 

Inconsistent policy communication leading to confusion and errors. 

Difficulty in managing and updating the knowledge base efficiently. 

Project details

Expertise used:

AI and NLP integration

Team composition:

  • Developer
  • QA engineer 

Service provided:

  • Data engineering
  • AI transformation 

Why a local AI chatbot? 

At GreenM, we have recently found out that most of us would like to use ChatGPT, however, are concerned about data privacy due to the sensitive nature of data. But, rejecting how this tool is helpful makes no sense. Plus, there is a need for a solution to navigate daily workflow smoothly, effectively manage, and communicate company policies, etc. 

 

 

After brainstorming sessions, we have figured out that the way out is to deploy your own GPT (LLM) model locally in your infrastructure so that data never goes outside and stays safe within the company. So, in our case, the aim was to create privately an AI chatbot capable of handling numerous inquiries, providing precise information, and maintaining an up-to-date knowledge base.  

Technology challenges 

  • Ensuring real-time response accuracy for HR and security queries. 
  • Developing context-aware capabilities for personalized responses. 
  • Integrating the chatbot with existing business applications like Slack. 
  • Supporting multilingual communication to cater to a diverse workforce. 

Tech stack

Python
PyTorch
Hugging face
Miniconda
Nvidia Cuda

Solution delivered

The AI chatbot we developed transformed the internal communication system and offered a positive and user-friendly experience designed to streamline the administrative tasks workflow, HR, security policy communication, and knowledge management. Key features and capabilities include: 

High performance 

The chatbot handles up to 150 queries per second, ensuring swift and accurate responses to staff inquiries. This capability allows for efficient handling of a large volume of queries simultaneously. 

 

 

Slack integration 

By integrating the chatbot with Slack, staff can easily access information and ask questions directly within their preferred communication platform. This integration streamlines the process of obtaining necessary information. 

Real-time context awareness 

The chatbot is equipped with context-aware capabilities, ensuring that responses are tailored to the specific needs and queries of users. This feature enhances the relevance and accuracy of the information provided. 

 

 

Customizable interface 

The chatbot interface can be customized to reflect the branding and visual identity, ensuring consistency and familiarity for users.

Task automation 

Routine administration, HR and security requests are automated, allowing staff to quickly obtain the information they need without manual intervention. This automation saves time for staff, allowing them to focus on what matters most.

 

 

Role-based management 

Different user roles, including HR administrators and general staff, are provided with tailored permissions and functionalities. This ensures that each user can efficiently perform their tasks and access relevant information. 

How it works

We picked RAG (Retrieval-Augmented Generation) to integrate our knowledge base, handbook, security and HR policies and procedures into the chatbot.

Having received a query, it searches for the most relevant portion in database that we integrated through cosine similarity. Then, an LLM (large language model) provides its response to the request based on the question and information parts, which the user gives at the first step. 

This ensures that reliable answers are given to questions.

What our colleagues say

Once the chatbot has been launched, I’ve gotten rid of the burden of answering teammates about the company policies. This tool has become kind of a relief.

Tetiana Kravtsova, HR Department 

 

“It helps me in getting an extract of this or that information out of the company docs. And it takes me very little time and effort. Earlier, looking for the details I needed in the documents got me distracted from my primary duties for a while. Now, it’s not the case.”

Oleksandr Serikov, BizDev Department

Local AI chatbot “triple S” advantages

Security

Private deployment of the chatbot in your environment eliminates risks associated with data breaches, contributing to compliance. 

Scalability

The cloud-based infrastructure allows handling numerous queries and users, ensuring consistent performance as the company grows. 

Staff engagement

The chatbot levels up staff involvement by providing instant, accurate responses and facilitating easier access to information.

It might be AI chatbot for healthcare or your domain

The potential of a tool like this is high. It could be a helper in healthcare to streamline patient inquiries, appointment scheduling, in e-commerce to improve the sales process, in education to assist students and staff with administrative queries, course information, and policy details, or in your use case. 

 

Alexey Litvin

CEO

 

 

Learn our clients’ experience

Copyright © 2024 GreenM, Inc. All rights reserved.