ServiceLater

Inspiration

ServiceLater exists because ServiceNow customers needed a clearer way to surface accelerator priorities. We built a platform that uses AI to match requests to accelerators and recommend new ones.

What it does

ServiceLater analyzes customer needs and recommends accelerators with a Pattern Analysis Agent that surfaces trends and an Accelerator Recommendation Agent that proposes new accelerators with justification. The React UI lets teams review requests, spot gaps, and use AI-powered insights.

How we built it

We started with a React app integrating Google Gemini for classification, matching, and analysis. We then migrated to Google ADK with a multi-agent Python backend (Parser, Classifier, Matcher, Pattern, Recommender), a FastAPI server bridging frontend and backend, and achieved about an 80% code reduction, all on React, Tailwind CSS, JavaScript Google ADK, FastAPI, and Gemini.

Challenges we ran into

We initially hit issues with manual orchestration of agents, Gemini 10 RPM rate limits requiring 6-second delays, and complex error handling. Migrating to Google ADK removed manual coordination, added a built-in tool framework, and simplified the architecture.

Accomplishments that we're proud of

We cut code by about 80% using ADK, built a multi-agent setup where agents call other agents, integrated FastAPI with our React frontend, delivered a feature-complete dashboard with analytics, and achieved real-time gap analysis with recommendations.

What we learned

Google ADK simplifies multi-agent systems and avoids manual coordination. Using agents as tools enables more flexible workflows, and less code can yield more reliable systems. A phased migration strategy kept existing features working.

What's next for ServiceLater

Next steps include real-time data integrations instead of CSV imports, fine-tuning on ServiceNow domain data, adding dynamic web tools, webhook notifications for recommendations, forecasting analytics, and user workflows with approval and feedback loops for continuous improvement.