Tasks

Every AI application receives thousands of conversations, but understanding what users actually get done with AI is often difficult. Most organizations measure AI usage through conversations, active users, or requests processed. These metrics show adoption, but they do not explain which tasks users delegate to AI, which types of work are most common, or where AI is becoming part of day-to-day workflows. Without this visibility, it is difficult to:

• Understand what people actually use AI for
• Identify the tasks and workflows where AI is most widely used
• Compare how different teams use AI
• Understand where users struggle to complete work with AI

Tasks turns AI conversations into a structured view of the work users are trying to accomplish. For employee-facing agents, this means moving beyond what employees talk about and understanding what they actually asked AI to do, such as drafting content, researching information, analyzing data, or troubleshooting a problem.

​

How Tasks are identified

Unlike most Nebuly metrics, Tasks are not assigned at the individual interaction level. They are identified across segments of a conversation. This is because a task is often completed through several interactions rather than a single prompt and response. A user might start by asking AI to draft something, then refine it, add context, or request changes across multiple follow-up interactions. Nebuly groups these related interactions into a segment, representing the sequence of interactions that contribute to the same task. Within each segment:

• One interaction acts as the main interaction, defining the primary objective of the task.
• Subsequent interactions that continue working toward that objective are treated as follow-ups and are associated with the main interaction.
• When the user’s objective changes, a new segment can begin.

As a result, a single conversation can contain multiple segments and multiple tasks if the user moves from one objective to another.

​

The Tasks report

The Tasks report shows what users are doing with AI and how that work is distributed across your organization. For each task, Nebuly reports:

• Interactions: Total interactions associated with the task
• Users: Number of unique users performing the task
• Error rate: Percentage of interactions associated with the task that resulted in an AI failure

Use this table to identify the tasks employees perform most often and compare how they perform across interaction volume, users, and error rate. To investigate a task further, open it in Explore mode. From there, you can drill down into its underlying sub-tasks, inspect the conversations behind the data, and understand how users are carrying out that work. Read more about Nebuly’s Explore mode here.

​

Tasks over time

Tracks how task volume changes throughout the selected period. Use it to see which types of work are becoming more common, which are declining, and how employees’ use of AI changes over time.

​

Most typed keywords

Shows the most frequently used keywords across conversations associated with tasks. This can help you spot recurring terminology, tools, documents, processes, or business concepts that appear across the work employees delegate to AI.

​

User intent over time

Shows how different categories of user intent evolve over time. Use it to identify seasonal trends, changes in user behavior, and shifts in how users interact with your AI.

​

Main tasks performed

Shows the tasks employees perform most frequently with AI. Use this view to understand which types of work account for the largest share of AI usage and where AI is becoming part of regular workflows. It can also help you identify high-volume tasks worth investigating further for efficiency, quality, or automation opportunities.

​

Build your own reports

Task dimensions are also available throughout Nebuly’s reporting engine. You can combine Tasks with your existing metrics, filters, and groupings to analyze how AI is used across teams, projects, business units, or other user segments. For example, you can build reports to understand:

• Which tasks are performed most frequently with AI
• How the task mix differs across teams
• Which tasks are growing fastest over time
• Which tasks have the highest failure or abandonment rates
• Which models or AI applications are used for specific types of work
• How AI spend, proficiency, or productivity differs by task

By analyzing AI usage at the task level, you can see where AI is actually being used in day-to-day work, which workflows are working well, and where users still need support.