Summary: ◀▼
This article shares best practices for using analyst copilot in reports, dashboards, datasets, projects, and memories. You’ll learn to organize projects by team or use case, define access roles before sharing, build focused datasets with the right joins and fields, use memory summaries to spot trends, and save useful findings so future analyses start with shared context and consistent metric definitions.
Best practices for projects
Projects help you to scope data, organize content, and control who sees what. Before sharing a dashboard or inviting teammates, take some time to structure your projects.
Organize your projects around how your team works
Create separate projects for different groups such as Support, IT, HR, specific regions, or time-bound initiatives like a product launch or business review.
When a user asks a question within the project, analyst copilot surfaces the reports, dashboards, and memories they have access to, so the better you organize and name your assets and projects, the more relevant your answers will be. A well-scoped "Support AMER" project or a "Q3 Launch" project will return far more targeted results than a single catch-all project.
Assign roles before distributing projects
Projects use three access roles:
- View: Users can see content but cannot make changes
- Edit: Users can modify reports and dashboards within the project
- Admin: Users have full control, including managing other users' access
Users need at least View access to a project to see data in any dashboard contained in that project. If you share a dashboard before setting access, viewers will be unable to see it. Make sure to grant access before sharing dashboards.
Best practices for datasets
Agentic analytics includes prebuilt projects for Support, Copilot, and ITAM. Each one comes with automatically installed datasets, dashboards, and reports designed for common scenarios. This library of analytical apps will expand over time.
Before you create a custom dataset or build a new report, open a prebuilt app and explore it. This gives you a clear picture of what data is available and how the pieces connect, without the risk of missing something foundational. Once you understand the structure, clone any prebuilt content into your own project and customize it from there. This is known as forking data, meaning that analyst copilot creates a copy of the original content without altering the original.
Build datasets with your analysis goals in mind
When you are ready to go beyond prebuilt content, dataset builder lets you join tables and shape a data model that reflects how your service operation actually runs.
Define your joins deliberately
When joining tables (for example, tickets and a returns custom object), be explicit about the relationship you are modeling. Poorly defined joins can produce inflated row counts or misleading aggregates that are hard to catch later.
Only include the fields you need
Select only the columns that are relevant to your analysis goal. A focused dataset runs reports more quickly and reduces errors when surfacing trends and comparisons. If you find yourself including fields "just in case," create a second dataset for that use case instead.
Best practices for reports and dashboards
Use memory summaries to interpret results
Once a report is built, memory summaries automatically flag statistically significant trends and anomalies. Read the summary before drawing conclusions from the raw chart. A spike that looks alarming may be within normal variation; a small but consistent shift may be the more important result.
See Creating reports in analyst copilot and Creating and managing analyst copilot dashboards.
Organize dashboards around your audience
Group related metrics on the same tab. Use separate tabs or dashboards for different audiences, for example, agents, supervisors, and leadership, rather than putting everything on a single tab.
Use interactive features to find root cause, not just to display data
Dashboard cross-filtering, drill-downs, and brush selection are designed for exploration. When you click a data point to filter the entire dashboard, or use brush selection to isolate a date range, that context carries into any deeper analysis you run. Use these interactions to move from seeing what happened to understanding why.
Save findings as you go
When you reach a result worth keeping, save your report before moving on. This will preserve the report as a memory that will be surfaced as a suggested starting point for future analyses.
Ask questions in plain language
You don't need SQL knowledge or analytics expertise to use analyst copilot. Open the analyst copilot tab in Analytics, type a prompt or question, and get a cited, data-backed answer with supporting charts. A few questions to start with:
- "Why is resolution time up this week?"
- "How is Copilot influencing agent productivity?"
- "How does first reply time vary by team?"
From there, follow analyst copilot's suggested next questions or click into cited memories to dig deeper. Each follow-up adds context and narrows toward root cause.
If you need advanced custom metrics or filters, switch to manual report mode. Use expressions and logic operators as needed for sophisticated analytics.
Get the most from your reports and dashboards
- Build and iterate on reports. Use drag-and-drop to add fields, adjust filters, and change visualization types. Save versions frequently.
- Use memories to combine previous analyses, narrate business stories, and build dashboards that give holistic views to different stakeholder groups.
- Enhance dashboards with narrative text, images, and links, use filters and cross filters for interactive exploration, and tailor presentation for your audience.
Best practices for memories
Memories are the mechanism by which agentic analytics improves the more you use it. Every saved memory becomes institutional knowledge that benefits the next person who asks a related question. Without memories, your team rebuilds the same analysis from scratch every time.
Rate memories to keep the most useful content surfaced
Analyst copilot surfaces existing memories alongside its narrative answers. Rating memories helps analyst copilot learn which memories are most relevant for your organization. Over time, this shapes the suggestions that appear first.
Use memories to standardize metric definitions across teams
Save key metric definitions (such as how your team calculates resolution time, or what counts as a reopen) as memories. This prevents definitions from drifting across teams or sessions and ensures that every analysis, report, and stakeholder conversation references the same numbers.
Share findings with narrate
When you have the reports and memories you need, select them and click Narrate. To select multiple memories, hold down the Shift key and select any relevant memories. Analyst copilot generates a plain-language summary of your findings that you can share with managers or leadership. This removes the need for stakeholders to interpret charts themselves and makes it easier to align around a shared understanding of your data.