Timely

Inspiration

Timely already helps more than 1,500 organizations across Mongolia manage attendance, schedules, leave requests, payroll, and workforce reporting. However, HR teams still spend significant time reviewing historical records, identifying patterns, and repeatedly answering the same operational questions.

We wanted to turn Timely’s workforce data into organizational memory—an intelligent system that remembers past decisions, understands company policies, and helps teams act proactively.

What it does

Timely is an agentic workforce-management platform that combines real-time attendance data with long-term AI memory.

The agent can:

Remember company policies, schedules, and previous HR decisions

Detect recurring lateness, overtime, absenteeism, and staffing patterns

Explain insights using relevant historical context

Answer natural-language questions about workforce operations

Recommend actions while keeping people responsible for final decisions

Instead of only showing reports, Timely helps organizations understand what is happening, why it may be happening, and what they can do next.

How we built it

CockroachDB provides a scalable and resilient data layer for employee events, organizational policies, conversation context, and long-term agent memory. Its distributed architecture supports Timely’s multi-tenant environment while keeping organizational data isolated.

AWS services power the agentic workflows and application infrastructure. The agent retrieves relevant memories, combines them with real-time workforce data, reasons about the request, and returns a contextual recommendation through Timely’s web and mobile experience.

Challenges we ran into

The biggest challenge was deciding what the agent should remember. Workforce data is sensitive, so memory must be relevant, explainable, tenant-specific, and securely isolated.

We also had to balance proactive automation with human oversight. Timely recommends actions and explains its reasoning, but important HR and payroll decisions remain under human control.

What we learned

Agentic memory is more valuable than a longer chat history. A useful business agent must remember structured events, policies, exceptions, and previous decisions—and retrieve only the right context at the right moment.

We also learned that AI creates the most value when it fits naturally into an existing workflow instead of becoming another separate tool.

What's next for Timely

Next, we plan to improve personalized memory retrieval, multilingual support, anomaly detection, and proactive workforce planning. Our goal is to make Timely a trusted operational copilot for organizations of every size.

Built With

• ai-agents
• amazon-api-gateway
• amazon-bedrock
• amazon-web-services
• aws-lambda
• cockroachdb
• node.js
• react
• react-native
• typescript
• vector