Research Preview: Agent

We have released an AI agent that analyzes data within the ontology, finds the causes, and proposes improvement options as a Research Preview.

Checking the data is only the beginning. Identifying causes, weighing alternatives, writing reports, and sharing them with the team. We believed that AI could help with this entire process.

For faster and more accurate decisions

In a company, everyone makes decisions. And these decisions take more time than you might think.

A decision is usually made in two steps. First, you look at the past. You review the data, check the metrics, and get alerts when something is off. Then, you look to the future. You think about what the cause is, how things will unfold, and what you should do.

Ontology, search, analysis, and alerts have helped you grasp the past faster and more accurately. The agent goes one step further. It greatly reduces the time spent reviewing the past and also points the way forward.

AI handles repetitive analysis and document work, while people review the results and provide feedback, focusing on the things only people can do.

What the agent can do

With the agent, you can assign tasks through chat, just like ChatGPT or Gemini. But this is different from an ordinary conversation. The agent works directly with the data connected to the Deskroom ontology and produces deliverables.

• Data retrieval and analysis: It can answer complex questions such as "Analyze why CS inquiries increased last month" or "Compare the churn rate by customer segment this quarter."
• Root cause analysis: When a metric fluctuates, it infers possible causes based on the data.
• Suggesting alternatives: Based on the analysis, it proposes directions for improvement. It does not stop at showing the current state but also offers ideas for the next action.
• Drafting documents: It automatically generates labor-intensive documents such as report drafts and team-wide sharing messages, grounded in the data.

The agent's design principles

The agent is designed to collaborate with people. This Research Preview embodies three core principles.

• Planning: When it receives a request, it builds its own plan for how to carry it out in stages. It breaks down even complex requests step by step.
• Clarification: If a request is ambiguous or more information is needed, it asks first. It begins work only after understanding things accurately.
• Execution: It carries out the work in the planned order and delivers the completed results. You can also transparently follow the intermediate steps.

Going forward, we plan to give the agent more hands and feet. So that AI handles repetitive work and people can focus on making the right decisions. We will support companies in making faster and more accurate decisions.