Quick Start

This page walks you through submitting your first query, reading the response, and understanding what the agent did to produce it.

Prerequisites

• Valid Copilot Assistant access. See Access and Trial.

• Assets configured in the portfolio, ideally with data already being ingested

Before you begin, confirm that Copilot Assistant is available for your tenant in Access and Trial, and review Improving Results.

Your First Query

The Copilot Assistant page is the central place to start new conversations, continue previous ones, and manage conversation history.

Copilot Assistant welcome page with the prompt input box and suggested prompts.

Starting a Conversation

1. In the top menubar, click on AI Assist [1]. ▷ The Copilot Assistant welcome page opens.

2. In the welcome page, enter a prompt or choose one of the suggestions below to start a new conversation [2] or continue a previous one from the conversation history on the left.

3. If you are happy with your prompt click on the Send Button [3]. ▷ The request is processed and intermediate progress updates will become visible.

A good first query is specific and easy to verify:

• "How many assets are in my portfolio?"

• "Show me all alarms with high priority."

• "What are the current temperatures in my buildings?"

• "Show me the average weekly energy usage for all buildings over the past six months."

Understand the Tool Calling Trace

As the Assistant processes a request, it calls tools to match the request to the asset structure, retrieve data, analyze results, and generate visualizations. The tool call trace shows how your request was interpreted and which processing steps were completed. Examining it is the fastest way to understand why a response looks the way it does and where to rephrase if it does not match your intent.

Example: Processing the query "Show me the average weekly energy usage of the past 6 months for all buildings"

Example tool calling trace for a weekly energy usage request.

For the query "Show me the average weekly energy usage for all buildings over the past six months", the Assistant called 4 Tools and invoked 2 subagents:

1. Resolving the location concept 'buildings' from the request under [E1]. ◈ It found a matching AssetType Building ◈ It examined several assets with that AssetType (e.g. several SmartOffice Buildings)

2. Resolving the measurement concept 'energy usage' [E2]. ◈ It uses the previously retrieved AssetType Building as a condition and only examines entities within those buildings ◈ As the request aims to analyse incoming data, only asset measurements with the subtype input are considered ◈ The request specifies a time range of the past 6 months, thus only assets which actively recorded data in that time range are retrieved ◈ Several asset types which measure energy usage and have asset instances that were active in the past 6 months were detected ◈ Results are shown as AssetType.attribute and also display the AttributeType and unit if any

3. Once the Assistant has determined the Asset Scope it invokes the specialized SQL Subagent to retrieve the corresponding data from the database [E3]. ◈ The SQL Subagent was able to query and return 255 rows of data - of which the main Assistant will see a 10 row preview

4. The assistant inspects and analyses the result set deterministically [E4]. ◈ It applied three operations on the whole database result to compute key results: overall mean energy usage, lowest average energy usage, highest average energy usage

5. It uses the computed metrics to create KPI cards that visualise these to the user [E5]

6. The assistant decides that Charts are appropriate based on the returned data in addition to the user request and invokes the Visualisation Subagent [E6]. ◈ The Visualisation Subagent determines to visualise the full 250 rows of data and not just a 10 row preview

   ◈ The Subagent created a line chart which will display the average weekly energy usage on the y-axis and the week on the x-axis. It also creates a series for each Building in the chart.

Reading the Response

Each response contains a textual explanation and can include result artifacts in the active chat window [A].

Copilot Assistant response with the conversation history, tool trace, data artifact, KPI cards, charts, and chat controls.

While each response includes a final textual response about what was retrieved and analysed it will also include several artifacts that show the underlying asset data and visualise the results. Once the Assistant processes your request it will make the result artifacts visible as soon as they are available. Each response can include:

Marker	Component	Description
[E]	Tool call trace	A log of which steps the agent took (expandable)
[H]	Data table	Raw query results
[I]	KPI cards	1-3 highlighted key numbers from the retrieved data
[K]	Charts	Auto-generated when data suits visualization (can be expanded in a Dialog [J])

Once the Assistant finishes the turn you have several options to continue:

1. Starting a fresh conversation by creating a new chat [C]. ▷ The initial Welcome Page will open.

2. Sending a follow-up Request [L] + [M]. ▷ A new turn is started - the Assistant still has the context of the previous turn.

If the Conversation Stream was interrupted at any point the Refresh Chat Button [G] can be used to fetch the latests updates without having to reload the page

The expanded Data & Query artifact is shown below. It contains the raw results, query editor, and export controls.

Expanded Data & Query artifact with raw results, query editor, and CSV export.

In this figure:

• [H1] identifies the expanded Data & Query artifact and its tabs.

• [H2] identifies the CSV export control.

• [H3] identifies the raw results table.

When a conversation becomes longer, switch to Tabs mode to navigate between requests.

The control above the chat window switches the viewing mode [D]. Tabs mode can be used to navigate between the requests in a longer conversation [D1].

Tabs mode for navigating between conversation turns.

Continue the Conversation

After a response, you can:

1. Create a new chat [C].

2. Send a focused follow-up question [L] using the composer [M]. Copilot Assistant retains the relevant context from the previous turn.

3. Open the Agent Debug side panel [F] to inspect detailed conversation and tool information.

If the conversation stream is interrupted, use the Refresh Chat button [G] to fetch the latest updates.

Additional Configuration

• Copilot Assistant Settings allow you to adjust the agent behavior [B].

• The Agent Debug side panel provides detailed information about the conversation and the results of called tools [F].

Tips

• Name the asset, location, or asset type when you know it.

• Include a time range for historical data.

• Include a unit when multiple measurements could match.

• Start with one clear question and refine it with follow-ups.

• Use the Prompting Guide when the result does not match your intent.