The client wanted to build a cost-effective and efficient BPA platform that helps enhance the productivity of business users by creating automated processes through a user-friendly interface. The vision was to build a platform similar to ChatGPT that leverages generative AI to automate complex business processes and workflows using English as code.
Our client is a pioneering technology startup based in California, leveraging the power of artificial intelligence to automate various enterprise business processes. The company's products and services empower business users to automate mundane tasks, drive innovation, personalize customer experiences, and enhance efficiency through plain English commands. Their SaaS platform integrates seamlessly with major ERP, CRM, and productivity solutions.
Our customer had successfully raised $4.5 million in a seed funding round to develop an easy-to-use automation platform that can extract valuable business insights from data obtained from various sources such as operations, sales, and other analytical and marketing tools.
Due to time and bandwidth constraints, they wanted to engage with an external engineering partner and quickly onboard skilled developers. They turned to Velotio because of our Web App Development, Data Engineering, Analytics, Automation, and Data Science expertise.
The client wanted to build a cost-effective and efficient BPA platform that helps enhance the productivity of business users by creating automated processes through a user-friendly interface.
The client aimed to remove historical barriers posed by legacy software systems, such as Salesforce, SAP, Workday, etc.
We opted for ChatGPT, like a generative AI chatbot interface, because it has several advantages over traditional interfaces. The vision was to build a platform that uses conversational AI to not only automate repetitive operations but also help redefine how business users interact with the systems.
Our collaboration with Velotio has been nothing short of exceptional. Their team seamlessly integrated with ours, quickly grasping our needs and providing valuable insights and suggestions. In just three weeks, they delivered a platform that far exceeded our expectations in terms of performance and flexibility. Their agile approach enabled us to launch the platform on time. What impressed us the most was their team's ability to communicate effectively, ensuring that everyone was on the same page throughout the entire development process. This level of transparency and collaboration significantly contributed to the project's succes
Our client aimed to bring the power of Generative AI to a search engine that runs English as Code to automate intricate business processes and workflows.
Velotio took on the challenge of building the custom platform within the expected timeline with a team of Frontend engineers, Backend engineers, UX designers, a product owner, and a QA.
After extensive research and brainstorming, we chalked out a plan to build a serverless system with the goal of making it extremely scalable and easy to use. Since the platform's vision was so closely related to the functionality of ChatGPT, we were anticipating a huge flow of user traffic to the platform as soon as it was released, so scalability was of utmost importance. To accomplish this goal, we leveraged AWS Lambda backends - AWS serverless resources - and utilized AWS S3 and AWS CloudFront services on the front end to ensure the system could smoothly scale to meet the demands of an unlimited number of global users on demand.
For an intuitive interface design, we built an innovative interaction with a system like ChatGPT. It allows users to take actions based on system-generated prompts. This easy-to-use interface enables users to seamlessly use everyday English as commands to automate business processes and workflows.
We identified the various business processes or workflows the chatbot would automate or assist with. We also identified the requirements for the chatbot interface, the types of queries it should handle, the level of personalization required, and the response time expectations. This helped guide the design and development process by defining the scope, goals, and objectives of the implementation. We also determined if the chatbot use case requires the advanced capabilities of GPT-4, such as natural language generation, sentiment analysis, or machine translation.
AWS offers a wide range of services that can be used to build and deploy our implemented platform, including AWS Lambda, Amazon S3, Amazon DynamoDB, Amazon Lex, and more. We worked with the client on cost and timelines to select the appropriate services based on the use case and requirements
There are several ways to integrate GPT-4 into the platform, such as using the GPT-4 API, building a custom integration with AWS Lambda, or using a pre-built integration tool like Hugging Face solutions like Inference Endpoints.
We worked with the client to decide if you want to build a custom GPT-4 model for their specific use case or use pre-built models available in the GPT-4 API. Pre-built models can save time and resources, but custom models can provide more specific and accurate responses required for the RPA. In a custom mode, we must train and fine-tune the model to improve its accuracy and relevance to the business process automation use case and understanding of the domain.
Integrating GPT-4 into the chatbot backend: Given the scaling requirements and ease of operationalizing the backend, we decided to use AWS Lambda serverless technology to integrate GPT-4 into the chatbot's backend. This enabled us to generate natural language responses and provide more personalized and sophisticated information.
Based on the specific requirements and use cases, choosing appropriate frontend and backend technologies was critical. We went with React for the front end based on our past experiences with similar scale and development complexity.
We decided to go with Python’s Flask web framework for the backend. Most of the ML and NLP libraries we planned to incorporate into the system were very mature, with great support, and readily available in the Python ecosystem.
We leveraged Amazon Lex to build the chatbot's natural language processing model, which could understand and respond to user input in a real-time conversational manner. Also, as mentioned above, AWS Lambda was used to host the chatbot's backend logic, which can process user input and perform actions based on the user's intent.
We used serverless technologies like Amazon API Gateway, AWS Lambda, AWS S3, and AWS Step functions to build scalable and cost-effective architecture. The architecture eliminates the need for managing and provisioning servers and scales automatically based on real-time demand.
We extensively used AWS security tools like Amazon Cognito and Identity and Access Management (IAM) to secure user data, implement authentication and authorization and restrict access to sensitive information.
Use AWS tools like AWS CloudWatch and AWS X-Ray to monitor and troubleshoot the chatbot's performance and identify issues.
Test, deploy, and optimize the GPT-4 chatbot: Testing the chatbot thoroughly was one of the essential steps for the success of this project. This would help ensure that it meets the requirements and is functioning correctly. When we deploy the chatbot to production, we continuously monitor its performance to ensure it delivers the intended benefits. We eventually also invested in optimizing the integration based on user feedback and analytics data.
To make the solution scalable further, we also proposed eventually using serverless architecture already in place, as mentioned earlier, along with AWS Elastic Beanstalk or AWS Fargate for containerization. These technologies provide horizontal scalability and high availability by automatically scaling resources as needed.
Served 120k global users within 72 hours of the launch with 99.8% uptime—cloaked 94.5% positive feedback on the responses from the platform.
Increased customer retention by 92% due to outcome accuracy, efficiency, and ease of use.
Built conversation and action plan integration support for 18 enterprise data tools and utilities like Salesforce, Workday, SAP, and more in less than a month.
Platform users reported increased productivity by 20x. Also, reduced the automation plan design and execution turnaround time from 60 mins average to 3 mins.