Know what every customer needs.

It overlaps with one, and that is not how it is built. A helpdesk organises messages into tickets. Resolivo models the situation behind the message — who the customer is, what they bought, what was promised, what is broken, who internally owns it and whether the relationship is holding up. If what you need is a shared inbox with ticket numbers, there are cheaper products that do exactly that well.

Roughly five to a hundred people, with anywhere from two to twenty handling customers. Below that a shared mailbox genuinely works. Above it you will start wanting things this deliberately does not do — workforce management, call-centre routing, enterprise ITSM. The sweet spot is a company that has outgrown the mailbox and does not want a contact centre.

No. You connect the shared mailbox they already write to, with its history. Asking customers to learn a new address is a migration cost paid entirely by them, and it is the most common reason a helpdesk rollout stalls.

An afternoon. Connect the mailbox, bring in your customers from whatever you have — a CRM, a CSV, a spreadsheet — point it at your existing help content, and say who handles what. Customer health needs a few weeks of history before it says anything honest, and the product shows it as insufficient rather than guessing.

The connected model underneath: a customer has contacts, conversations, cases, purchases, invoices, promises, entitlements and a health state, and each of those is a real record with a real relationship to the others. It is what lets the product answer a question nobody thought to record the answer to, such as which customers renewing this quarter have an unresolved complaint.

Only if you turn that on, only for categories you choose, and never for refunds, cancellations, contract changes, account access, complaints or anything involving sensitive data — those are excluded structurally rather than by a setting. The recommended starting point is suggestions only, and before you widen it you can see how it would have handled real past cases.

It answers from records and verified knowledge rather than from its own recollection. Facts like a renewal date or a payment status are read from fields with sources, every answer names the records behind it, and when it finds conflicting information it says so and escalates instead of picking a side. Confident invention about a real customer is the worst possible failure in this category and the design is arranged around preventing it.

No. Questions run with the identity of whoever is asking, through the same permission checks as every screen, so there is no separate AI permission model to drift out of step. In the customer portal it answers as that customer — their own account and published knowledge only.