Background – AI Onboarding Assistant
Keepr is a first-of-its-kind financial incentive platform that specializes in profit sharing, stock options and employee bonus plans.
Keepr is a subsidiary of Keeper Solutions. It is thus the perfect playground for the team at the Keeper Solutions AI Lab to explore AI opportunities and develop proof of concept solutions. In the past, the team has developed an AI assistant to help answer queries and schedule sales calls.
Keepr wanted to simplify the onboarding process, particularly for larger enterprises, where the complexities of internal profit sharing or stock options schemes can make it quite time-consuming for the internal team. As the number of companies using the rewards system grew, the team at Keepr wanted to explore the AI opportunities and see if an AI bot could be used to streamline the entire process.
Challenge – Simplifying The Onboarding Process
There are many elements involved in awarding profit shares or stock options to team members. This includes factors such as seniority within the organization, time at the company, bonuses accrued etc. While this task might be straightforward for smaller startups, it becomes significantly more complex when dealing with larger enterprises. The goal of the project was to train an AI assistant to understand, simplify, automate and then accelerate this complicated stock-allocation process.
The key challenge was integrating a chatbot system that could understand user intent and perform specific actions within the Keepr platform. The AI assistant needed to accurately interpret user prompts and execute corresponding tasks on the platform, all while minimizing AI “hallucinations” (instances where the AI generates incorrect or nonsensical responses).
Solution – An OpenAI Assistant
The solution involved developing a sophisticated onboarding assistant that was made up of a number of technologies working in tandem.
-
-
-
-
-
-
-
-
-
-
- OpenAI’s Assistants API was used as the core AI model that would receive user prompts (i.e. messages from the user), interpret the intent of those messages and respond in a way that it was programmed to do.
- The system was built around a FastAPI server that acted as a middleware, handling communication between the front-end user interface and the backend platform.
-
-
-
-
-
-
-
-
-
Our team at the AI Lab then established a logic flow to ensure that the onboarding assistant could receive prompts, understand user intent and respond accordingly. A well thought out logic flow would ensure seamless communication between the user interface, the AI and whatever backend system is required to perform the required task.
Here’s how the solution works:
-
-
-
-
-
-
-
-
-
-
- User Interaction: The user types a prompt into the front-end and submits it.
- Backend Communication: The front-end sends this prompt along with the user’s location to the FastAPI service.
- AI Processing: The FastAPI service relays the prompt to the OpenAI Assistant API, which processes the input based on its training and returns a structured response in JSON format.
- Action Execution: The FastAPI service either directly responds to the user with the AI’s answer or interacts with the Keepr Platform’s backend (built with Django) to execute an action, such as creating rewards tiers based on company roles.
- User Feedback: The front-end component updates to reflect the changes, providing immediate feedback to the user.
-
-
-
-
-
-
-
-
-
Below is the full tech stack:
-
-
-
-
-
-
-
-
-
-
- Frontend: Custom React component
- Middleware Backend: Python FastAPI
- AI: OpenAI Assistant API
- Server: Google Cloud Platform/Cloud Run Service
- Keepr Backend: Django
-
-
-
-
-
-
-
-
-
Results – Can We Use AI For This?
After experimenting with multiple technologies including Intercom’s Fin and ChatterBot, the team eventually uncovered a solution, listed above, that has completely transformed the onboarding process within the Keepr financial incentive platform.
The AI assistant has greatly improved the user experience, particularly for super users who manage their own onboarding process. By making the onboarding process smoother and more intuitive, the AI assistant has increased user satisfaction, and made life easier for all involved.
A simple exploratory project, where a team wanted to see if they could use AI to remove a time-consuming and often arduous task, has made a lasting impact. The project once again, highlighting the many opportunities that exist when you continually ask ‘Can we use an AI for this?’