Find your way around the workspace

Account, workspace and website#

Your account identifies the person who signs in. A workspace contains the business's websites, sources, conversations, usage and team membership. A website identifies one customer-facing domain and its installation.

If your account belongs to more than one workspace, the top bar shows a workspace selector. Switching it changes the data and configuration you are viewing. Knowledge is scoped either to one website or to the whole workspace, so deliberately choose the right scope before saving an entry.

Read Overview correctly#

Overview shows whether access is active, the current usage period and counts for websites, conversations and processed Knowledge pages. Websites includes an installed count; installation and successful Knowledge collection are separate checks.

The support activity chart can show seven or 30 days, and starting-page bars identify where conversations began. These are observed support counts, not proof that a particular reply caused a sale. If there is no active access period, activate your issued access code or choose a plan before adding a website.

The main working areas#

Plan-dependent areas#

Business and Scale enable multiple websites, team invitations, human escalation and recurring-pattern insights. Their navigation can include Escalations and Setup. Scale enables journey analytics and custom widget identity or branding removal. Core retains the same assistant intelligence with a smaller capacity and feature set.

A missing control can reflect the selected workspace's plan or your role, not a browser fault. Full workspace management requires an active owner or admin membership. See Team access and roles before inviting someone to configure sources or integrations.

Personal controls and recovery#

The Light or Dark control changes the portal theme stored in your browser. It does not change a website widget's theme; use that website's Appearance tab for customer-facing settings.

The account menu opens Account settings and Billing & usage. If a view cannot load, confirm the selected workspace and your role, then reload. Do not recreate a website or make a second purchase simply to repair a temporary loading failure. Keep the visible error and the affected domain for support.

Worked example: more activity without a billing discrepancy#

An owner sees more support activity in a thirty-day chart than they expected from a seven-day review. First they compare the selected reporting range rather than treating the change as duplicated traffic. They then use Billing for the active allowance period, and Conversations for retained examples of what was asked. These screens answer related but different questions.

If an older thread is no longer available, its absence does not refund used capacity or prove that the chart is incorrect. If a customer discussed several issues in one long exchange, the visible thread count is not necessarily the same as the units charged under the conversation rules. The owner records the range and measure before asking support to investigate an actual unexplained difference.

Make a configuration handover reproducible#

Ask the next maintainer to select the intended workspace, find the relevant website and identify its current source or integration. Have them perform a permitted test with their own account instead of merely reading a screenshot from the previous owner. The change record below describes what evidence to keep.

If they cannot reach a management area, check membership and plan before repeating the business change. If they can reach it but see another domain's settings, correct selection first. A handover is complete when the responsible person can locate and verify the actual configuration, not when somebody forwards an owner password or a private installation credential.

Follow a customer problem through the workspace#

Suppose a customer says the assistant quoted last year's delivery charge. Start by identifying the website and the retained conversation, if it is still available. Read the actual question and answer. A customer may have asked about another destination, a special product or a different business location. That context determines which policy you should compare; the headline complaint alone is not enough to identify a source defect.

Next open Knowledge for that website and inspect the relevant stored source. Compare its contents with the business's current public page. If the stored source is old, refresh or correct it using the supported source flow. If both pages disagree with an approved internal policy, the business must first decide which policy is authoritative. An assistant cannot resolve a commercial contradiction by choosing the most recent-looking sentence. If the stored source is correct but the answer mishandles an exception, use a precise Answering guidance correction.

Finally use Private test for the corrected question and an adjacent exception, then check the installed page if page context matters. Record the source changed, the expected answer and the observed result. This route moves from customer evidence to stored information to a focused verification. It is more effective than changing the theme, replacing the installation token or upgrading a plan in the hope that a different answer appears. Those controls own different parts of the system.

Read totals without inventing conclusions#

Overview brings several operational measures together. Used conversation units belong to the active billing pool. Retained transcript counts depend on how many threads remain available under retention. Knowledge document counts describe stored extracted documents. These measures have different denominators and lifecycles. A long chat can consume additional billable answer groups, while an expired transcript can disappear from review without refunding previously used capacity.

Activity for a seven-day or 30-day range is not the same as the current subscription usage period. Select the range that matches your question and note it when comparing results. If you compare last week's activity with a monthly allowance, say what each number represents. Changing the range does not reset usage or extend transcript retention. Likewise, a website with more processed documents is not automatically providing more accurate answers; inspect the relevant source and a representative conversation.

Analytics can report observed patterns or supplied outcomes, but a chart alone does not establish that the assistant caused a sale. A revenue-associated event comes from the configured integration, not from reading a conversation and guessing a payment. Use reports to choose investigations and content improvements. For commercial reporting, reconcile the relevant event with the system that owns the outcome. This keeps the workspace useful as an operational tool without turning its summaries into claims they cannot support.

Keep a useful maintenance and handover record#

Create a short record for changes that affect customers. Include the selected workspace, public website, date, responsible person, setting or source changed, and the acceptance check performed. Do not store usable installation credentials, invitation links, customer verification codes or copied payment details in a broadly shared document. References and configuration descriptions are enough to identify most maintenance work.

For a source edit, record the approved business policy and the representative question checked afterward. For a website release, record the active integration method and important public routes that were tested. For a handoff change, record the destination mailbox and whether an authorized test request reached it. For a billing decision, record the chosen plan or top-up setting separately from the website release. Keeping these outcomes separate prevents a colleague from interpreting one successful check as proof that every component works.

Use this record when an issue reappears. If a launcher disappears after a theme change, you can compare the previously working template with the new publication. If an answer becomes inconsistent after a shared-source edit, you can identify which websites were intended to inherit it. If an outgoing teammate owned the mailbox or provider endpoint, the record identifies an external dependency needing reassignment. Finish the handover by asking the replacement maintainer to perform a permitted check with their own account. A list of settings without an observed result is incomplete evidence of a working release.