Case study
AI Assistant for Effortless and Reliable Time Logging
Building an AI microservice and AI agents to automate time tracking via natural language
The Client - eBillity - offers a robust time-tracking and billing software trusted by thousands of professionals across industries like legal firms, consulting, and trade services. Their platform streamlines time logging through flexible web and mobile apps, with seamless integrations to major accounting systems such as QuickBooks and Xero.
To further enhance efficiency for users managing complex projects and diverse billing rules, eBillity sought an AI development partner to build a standalone, LLM-based AI API capable of interpreting natural language prompts, automating time entry, handling software integrations, and applying validation rules to help teams save time and reduce manual errors while maintaining the high standards their clients expect.
SoftKraft established a highly collaborative service process, working in close partnership with the Client's Product and QA Leaders, as well as their Backend Developers. Daily technical meetings focused on reviewing feedback on AI agent changes and test cases, which helped progressively refine the next steps.
To maintain strategic alignment and report on broader progress, the team conducted weekly updates via Google Meet. These sessions reviewed accomplishments from the previous week and outlined concrete plans for the week ahead, ensuring all stakeholders stayed synchronized on the project's status and trajectory.
SoftKraft successfully engineered and delivered a sophisticated, standalone AI API. The project, structured across three sprints, focused on building a robust microservice that intelligently interprets natural language prompts and enables reliable time logging. The solution leverages a powerful combination of FastAPI for the backend and LangGraph for creating supervised AI agents.
The final solution adeptly manages complex user interactions, from basic time entries ("Log 3 hours on project Alpha") to multi-step conversations that clarify missing information, such as a client or project name. For longer chats involving extended back-and-forth to resolve ambiguities, optimizations were implemented to minimize resource and token consumption, ensuring efficient performance without compromising accuracy.
SoftKraft delivered an AI API capable of handling intricate validation rules and field dependencies tied to various software integrations. The AI API now seamlessly manages time entries, runs active timers, and ensures all data passed to the main API is accurate and complete.
