How AI report counts and visibility work
Report counts from Ask Calm and Agent MCP are complete: the reporting tools aggregate every matching record the caller is allowed to read, not a sample. Everything an AI answer shows still follows your normal access rules, masking, and privacy limits. For how each surface connects and runs, see How Ask Calm and Agent MCP work.
Complete counts versus samples
Headline totals in the reporting metrics are complete. They aggregate every matching record the caller may read, so a ticket or work-order count covers the whole permitted population, not a slice of it. Search results, attention lists, and composed reports are different: they are compact samples, and they must not be substituted for the reporting aggregates.
Composed results carry a countCompleteness flag. A value of lower_bound means the count was taken after filtering and deduplicating capped reads, so it can understate the true permitted population, including when the reported count is zero. A value of exact means every source record was read, although the returned rows may still be capped. Neither flag turns a sample into a complete monthly report: for full counts, use the reporting tools covered in Use Ask Calm for ticket and work-order reports and Ticket and work-order reporting with Agent MCP.
A lower_bound count of zero does not mean the permitted population is empty. It means the capped sample contained nothing, and matching records may still exist outside it.
Available versus unavailable
An unavailable result means the module or population itself cannot be read with your access. An available result with zero matches means the opposite: the caller was allowed to read the population, and nothing in it matched. The two answers are distinct by design.
When only part of a result is blocked, a breakdownUnavailable explanation says which breakdown is unavailable while the totals remain available. A refused individual record is indistinguishable from a record that does not exist: the answer cannot tell you which of the two you hit.
Your normal access applies
AI answers follow the same visibility rules as the app: site, category, location, asset, company, personnel, and inherited-site access all apply unchanged. Ask Calm uses the signed-in user's access on the current site. Agent MCP uses the sites granted to the connection, then the connected person's remaining access on top of that.
Service Desk visibility follows your module role. A Service Desk Member sees only the tickets they created or were assigned, while Managers and Admins see all issues and requests on the site. See Service Desk roles and access for the full role breakdown.
Masked and hidden information
Anonymous reporters stay anonymous in AI answers, and contact details stay omitted. When a linked location or asset is hidden from your permissions, its name is masked. Readable parent records still appear in headline totals even when a linked location is hidden.
Selecting a location or asset filter requires access to that location or asset and a matching organisation or site record.
Module availability
The module-specific reporting tools are advertised only when the relevant Service Desk or Work Orders capability is available. When a tool is available, its results are always subject to normal site and record access.
Time and SLA
Durations in AI reports are calendar (elapsed) hours, not working hours. We do not provide native SLA targets, paused-time policy, working calendars, or SLA attainment. If you track SLAs, apply your own rules to the available timestamps and recorded transitions.
Privacy
Before Ask Calm sends site information to the AI provider, the consent step states: "Answers are produced by sending relevant site information to Calm's AI provider. Models are not trained on your organisation's data." For the underlying detail on storage, encryption, and AI routing, see Data storage, access, and AI privacy.