Last updated:

RAGi AI agent demo: from handover notes to a reconciliation answer

This 60-second video shows RAGi's AI agent handling a month-end reconciliation. One extra box turned up in June because both the Taipei and the Kaohsiung warehouse counted it: which one should it belong to? The AI agent finds a colleague's handover notes in the company knowledge base on its own, reads the whole document, checks the delivery date, applies the rule in the notes to count the June box for Taipei only, and writes a reconciliation note with its sources; once the draft is approved, the final report is ready to download. The scenario is illustrative.

Chapters

  • 0:00One box too many: which warehouse does it belong to?
  • 0:20The AI agent searches, reads and checks on its own
  • 0:46Download the final report; RAGi can also run in your own server room

What the AI agent does in the video, step by step

  1. Understand the question: decide whether the box belongs to the Taipei or the Kaohsiung warehouse.
  2. Search the knowledge base: find the handover notes a former colleague left, without anyone digging for them.
  3. Read the whole document, then check: read the notes in full, then check the box's delivery date. The notes say stock counts for Kaohsiung only from the day Kaohsiung checks it in, so the June box belongs to Taipei alone.
  4. Write the note with sources: the reconciliation note lists its cited sources underneath so a person can check them.
  5. Produce the report after human review: a person approves the draft, then the final report is downloaded.

The RAGi capabilities behind the video

RAGi's AI agent breaks the work into steps: searching the knowledge base, reading documents, checking data and writing the note, with cited sources so people can verify them, and key steps can require human review. RAGi supports on-premises deployment, keeping model inference, documents and indexes inside your own environment.

Learn More About RAGi Enterprise AI Agent Platform →

Transcript

The dialogue in the video is in Chinese; below is a translation.

0:00Hey! June's books have one box too many!

0:03Look, the Taipei warehouse counted this box, and so did Kaohsiung!

0:06It reached Kaohsiung at the end of June but was only checked in on July 1.

0:10So which one is it? I asked RAGi to look into it.

0:12We need the handover notes our old colleague left.

0:15I'll find them for you!

0:16No need.

0:18Look, it has already found them.

0:20It went through the company's files itself, read the whole notebook, then checked this box's delivery date.

0:27The notes say stock counts for Kaohsiung only from the day Kaohsiung checks it in. So this June box counts for Taipei only.

0:34It even wrote the reconciliation note for me, with the sources underneath.

0:40What? It used to take me half a day just to find that notebook.

0:44Hey, now you're the one teaching me!

0:46Once the draft is approved, you can download the final report.

0:51RAGi can also be installed in the company's own server room.

0:55Hand it to RAGi, and the AI agent finishes the whole job.

FAQ

How does the AI agent in the video find the answer?

It searches the company knowledge base for relevant documents and finds the handover notes a colleague left behind. After reading the notes in full, it checks the box's delivery date against them, applies the rule in the notes to decide that the box belongs to the Taipei warehouse, and writes it up as a reconciliation note with its sources.

Can RAGi be installed in our own server room?

Yes. RAGi can be deployed on premises or in the cloud. With an on-premises deployment, model inference, documents and indexes stay inside your own environment, which suits finance, manufacturing and government.

Another demo: live captions and translation in a multilingual meeting →