Getting Started with Enterpret Agent
Enterpret Agent helps you investigate customer feedback, explore patterns, and create useful outputs such as reports, presentations, and CSVs. Start with one question, refine it through follow-ups, and review the supporting feedback before using the findings.
Your first path: Ask a focused question → explore the answer → check the evidence → share the result or make it recurring.
If you previously used Wisdom, Agent extends that analysis workflow with Sessions, tools, and file outputs. Features and navigation depend on your workspace.
1. Start your first Session
Open a new Session. Choose Ask Agent, or open Agent → New session, depending on your navigation. If Agent is unavailable, contact your Enterpret team.
Choose one topic. Start with a product area, customer need, or issue you want to understand. For feedback analysis, the relevant sources must already be connected and their records available in Enterpret.
Ask in plain language. Include your sources, dates, customer segment, and desired output where relevant. You do not need a Skill to ask your first question.
Follow the work. Review the Session's progress and tool activity. Respond to questions or approval requests, then use follow-ups to refine the result.
Not sure what feedback is available? Start with Feedback Sources. If expected records are missing, use the data-coverage troubleshooting guide.
Try a first question, then build on it
Replace the bracketed text with your own product area and dates:
Summarize the main onboarding issues in [support source] from [start date] to [end date]. Group the recurring themes, count distinct feedback records for each, and include supporting record links. State any coverage limits.
Then ask one follow-up at a time:
“Show me the source examples behind the largest theme.”
“Break that down by customer segment, where that information is available.”
“Compare it with the previous period using the same sources and filters.”
“Turn this into a short team update with links to the evidence.”
Other starting points include “What feature requests appear most often in enterprise-customer feedback this quarter?” or “How has checkout feedback changed since our release?” Specify what should be counted and which dates or release you mean.
For more examples covering overviews, investigations, reports, and comparisons, use the feedback analysis prompt library. Restate important constraints if you change the scope during a conversation.
2. Choose the help your question needs
When should I use a Skill instead of simply asking a question?
A question tells Agent what you want to find out. A Skill supplies a reusable method or format for a particular job. You can use them together.
| Your goal | A good starting point |
|---|---|
| Explore a question or investigate something new. | Ask Agent directly and refine the answer with follow-ups. |
| Follow a repeatable method, such as your team's VoC readout. | Select an appropriate available Skill, then specify the topic, data scope, dates, and output. |
| Repeat a report on a schedule. | First refine the request in a Session, then configure an automation. A Skill alone does not schedule work. |
For example, ask “What onboarding issues appeared last month?” for an individual investigation. For a recurring team readout, a Skill can package the comparison method, evidence checks, and report structure you want to reuse.
Type / to select a Skill from the available list, or use the Skills control when offered. You can also describe the workflow in plain language. Select a particular Skill explicitly when you want to request that method. A Skill guides the work; it does not guarantee identical or correct results. See Skills for examples, scopes, and authoring.
When do I need context, connectors, or attachments?
The plus menu can offer Attachments, Skills, Connectors, and Manage rules. Choose what the task needs:
| What you need | Where to learn more |
|---|---|
| Product documentation, terminology, or playbooks that help explain your business. | Knowledge Base |
| Standing instructions or shared definitions; saved personal preferences. | Rules and Agent Memory |
| Additional context or a supported action in an external tool. | Connectors |
| A file for this investigation. | Add supported files through Attachments. Follow the formats and limits shown; see the Sessions reference. |
A feedback-source integration and an Agent connector serve different purposes. Connecting a tool for Agent does not automatically ingest its content as feedback. Connector tools depend on the authorized account, external-service access, and workspace permissions.
3. Read the answer and check the evidence
Use the answer to identify what to explore next. Before using a finding in a decision or sharing it with your team:
Check the scope. Do the sources, dates, topic, and customer segment match your question?
Check the measure. Feedback records, users, and accounts are different counts. Check percentage denominators and whether themes overlap before adding numbers together.
Open supporting records. Check that quotes and examples support the claim and preserve the customer's context.
Separate patterns from explanations. A recurring complaint can identify something worth investigating; it does not by itself prove the technical cause.
Generated answers and summaries can be incomplete or incorrect. Ask “What scope did you search?”, “What is being counted?”, or “Which parts are hypotheses?” when the answer is unclear. Use the answer-verification checklist for more detail.
4. Create and share an output
Ask for the deliverable you need, such as “Create a short report with the main themes, counts, and supporting links.” Where supported, Agent can generate a PDF report, slide deck, HTML dashboard, chart, or CSV.
Open the actual file to check its numbers, quotes, citations, and presentation. Check file visibility and recipient access before sharing. See Artifacts for supported outputs and sharing.
For a page that stays current, ask for a live HTML dashboard or tracker and give a time window, for example, “Build a live HTML dashboard of checkout complaints from the last 30 days.” Open it in Enterpret to refresh supported data. See Live Artifacts for refresh behavior and current limitations.
5. Return to your work or make it recurring
Find earlier work or start from feedback
Sessions are saved automatically. Open Sessions and use Search sessions to reopen earlier work and add follow-ups where allowed. Check the Share control before distributing a Session link. You can also start from a feedback record using Ask Agent when offered.
On mobile, open the navigation menu to find Sessions. Close a generated-file preview to return to the conversation if it covers the chat.
From a Session to recurring reporting
For a new automation that needs a chat upload, ask Agent to use the specific file. Review the resulting instructions and saved attachments. To change files on an existing automation, use its editor and Save. Follow the reference-file setup guide for upload steps, supported formats, limits, and previews.
Once you have a useful reporting request, configure an automation with the prompt, scope, schedule, timezone, destination, and any required Skills or connectors. Review the saved configuration, then use Run now where available to check its output and delivery.
A request in a Session does not by itself confirm that a schedule is active. See What are Agent Automations? and Scheduled Agent Automations.
Common beginner questions
Do I need to write a perfect prompt?
No. Begin with one focused question and refine it. Naming the sources, dates, customer segment, and output helps make the request clearer. Use the prompt library as a starting point and adapt its placeholders.
Does Agent automatically know our internal terminology?
Give it the relevant context rather than assuming it knows every internal definition. Keep product references in your Knowledge Base, use Rules for standing definitions, and specify important constraints in your request. Check its interpretation in the answer.
Why might two answers show different numbers?
Compare the sources, date windows, filters, measured object, denominators, and available data. Ask Agent to explain the scope and calculation behind each result. A difference needs investigation; it does not establish which result is correct.
What if there are no results or something seems missing?
Check the searched scope and whether the expected source data and fields are available. An empty answer does not prove that no customer reported the issue. Use the missing-data checklist.
Which model, speed, or thinking setting should I use?
Choose among the models and thinking levels offered in your Session. Where Fast mode is available, it uses priority processing for faster responses at a higher cost; it is separate from choosing a model or thinking level. See the Sessions reference for available controls, approvals, and limits.
Beyond Enterpret Agent
When you want to explore other ways of working with Enterpret:
Quantify: build charts and comparisons with explicit filters and measures. Get started with Quantify.
Dashboards: bring saved Quantify charts together for an ongoing team view. Get started with Dashboards.
Enterpret MCP: access supported Enterpret tools from a compatible external AI client. It does not automatically inherit your Agent Session’s Skills, Rules, or connector configuration. Set up Enterpret MCP.
For the broader platform overview, see Introduction to Enterpret.