AI Assistant for Faster Energy Insights

Overview

Hotovo developed DIMI, an AI assistant for Energy Aspects designed to make it easier and faster for users to access insights across research and time-series data.

The solution brings information previously spread across multiple applications into a single AI-powered interface. DIMI delivers precise answers and charts while intelligently routing each request to the appropriate module based on the type of information the user is looking for.

The Challenge

Energy Aspects needed a simpler way for customers to consume data distributed across multiple applications.

A key challenge was creating a powerful search experience capable of connecting insights across different types of content, including:

• research publications,
• time-series data,
• API documentation,
• user-specific information.

The system also needed to determine which AI agent should handle each individual query, ensuring that users were directed to the right source of information and received relevant, accurate responses.

The Solution

Hotovo developed DIMI, an AI assistant that provides answers and charts based on Energy Aspects’ research and data while routing each request to the appropriate module.

Search Module

To provide the AI agent with the most relevant results, Hotovo defined specific characteristics for each type of content, including research publications, time-series data and API documentation.

Based on these characteristics, the team developed a tailored agentic search module capable of handling the requirements of different content types.

An extensive evaluation set covering both common and advanced queries was created to validate the quality of the search experience.

Routing Module

Selecting the appropriate agent to handle each query was another crucial part of the agentic flow.

The routing mechanism was developed and refined using real user queries. Its accuracy was then evaluated using an additional extensive evaluation set created with input from domain experts and user feedback.

Barriers & Challenges

The project required the team to solve two interconnected challenges: retrieving the most relevant information from different types of content and reliably determining which specialised agent should process each request.

Because research publications, time-series data and API documentation have different characteristics, a single generic search approach was not sufficient. The search experience therefore needed to be tailored to each content type.

At the same time, the routing mechanism needed to reliably interpret real user queries and direct them to the appropriate module.

To address these challenges, Hotovo used extensive evaluation sets, domain expertise and real user feedback to continuously validate and refine both search and routing quality.

Results

DIMI unified previously scattered data within a single AI assistant, allowing customers to access and consume content in one place.

Key outcomes include:

• advanced search across research, time-series data and API documentation,
• tailored agentic search for different content types,
• intelligent routing to the appropriate agent for each query,
• filtering by commodity and time to provide more precise answers,
• instant chart generation with the option to jump directly to Data Explorer,
• license-aware access control, ensuring users only see content they are permitted to access,
• faster analysis and an improved user experience,
• higher client satisfaction,
• quality validated through evaluation sets and real user feedback, resulting in over 90% positive response ratings.