Know exactly what Cobotic does in your PSA

What Cobotic reads

Cobotic works through its own user in your PSA. You create it as an API-only user with only the access the AI workflows need, and it needs no extra PSA licence.

• Your PSA tells Cobotic when a ticket is logged, reassigned or resolved. Cobotic then reads that ticket.
• On each ticket it reads the subject, description, notes, category, priority, team and status, and the contact and company the ticket belongs to.
• It reads your own lists of categories, priorities, teams, ticket types and statuses, so every value it sets comes from your set-up.
• It searches your knowledge base for fixes your team has already written.
• It sees only the tickets in the teams you add it to.
• At setup it reads your last 90 days of tickets to learn how your desk files, routes and prioritises work.
• Before any ticket text goes to an AI model, Cobotic swaps names, email addresses, phone numbers and passwords for placeholders.

What Cobotic writes

Every change is made by the Cobotic user, so your PSA shows exactly who made it.

• Ticket fields. Category, priority, team and ticket type, set only to values already on your own lists.
• Internal notes. The ticket summary, the suggested fix, the handover note, alarm details, quality check results, and any category change with the old category recorded. Every note is internal and hidden from your users.
• Tags. Flags such as Cobotic alarm, Cobotic spam and Cobotic unhappy customer, so your techs see them in the ticket list.
• Knowledge articles. Saved as unpublished drafts. They stay unpublished until you approve them. Where a PSA cannot hold a draft, the article comes to you by email instead.
• New categories. Proposed in your monthly report. Cobotic never creates a category.
• Alerts. Breaches, outages, unhappy customers and possible scope creep go by email to the people you name.
• One connection test. When you connect, Cobotic makes one test ticket, adds a note and closes it, to prove its access works. You can delete it.

Cobotic never deletes or closes your tickets, never publishes a knowledge article and never emails your users unless you switch that on.

Your users hear from Cobotic only when you say so

Two AI workflows can write to your users: Instant Details Request and Self-Fix Reply. Both start with user emails off. Until you ask us to switch them on, each one writes its message as an internal draft note for your tech to send.

When you do switch them on:

• They never email a user on a top priority ticket, a breach or an outage.
• Self-Fix Reply sends only when its confidence clears a bar set from your own past tickets.
• The email goes out through your PSA, on the ticket, like any other reply from your desk.
• Instant Details Request moves the ticket to your waiting status. It is the only status change Cobotic makes.

The Test Drive key is read-only

The Test Drive connects to your PSA with a key that can only read tickets.

• The setup guide gives the key read access to tickets and no access to anything else.
• When Cobotic signs in, it asks your PSA for read permission only, and it makes read calls only. Nothing is written to your PSA.
• It reads your newest 150 closed tickets.
• The key travels with each read, is used for that run only and is never saved.
• Your browser removes contact names, email addresses, phone numbers, IP addresses and signature blocks before anything goes to an AI model. You see exactly what we receive, and you can delete anything else before you press Run.
• The Test Drive runs on its own AI account, kept apart from the one that serves paying clients.
• When the run finishes, delete the key or set it to inactive in your PSA.

Rather not share a key at all? Upload a ticket export instead and get the same report.

You stay in control of every AI workflow

• Tested before it touches a ticket. Before go-live, every AI workflow runs on your past tickets with nothing written to your PSA, and each one gets a score.
• A minimum to pass. An AI workflow that scores below its minimum stays off unless you switch it on. Red Alert Flag must catch all 10 planted test alarms before it goes on. Scope Creep Catcher must catch 9 of 10.
• Your call. You approve which AI workflows go on, and you can switch any of them off from your dashboard at any time.
• Drafts wait for you. Knowledge articles stay unpublished until you approve them, and new categories wait in your monthly report.
• Every change on record. Cobotic records which AI workflow made each change, and your PSA's own history shows every action the Cobotic user takes.
• Nothing half done. If an AI model call fails or your PSA refuses the key, Cobotic writes nothing to that ticket and we are told straight away.

Where it runs and who can see it

Cobotic runs on Cloudflare. Its AI models come from Anthropic, through Anthropic's API. Anthropic processes ticket text outside Australia, and it may be processed in the United States or other countries.

• Your PSA key is encrypted with a master key held apart from the database, and it is decrypted only in memory for each call. No screen or report shows it.
• Your key is locked to your account. Copied anywhere else, it will not decrypt.
• Our admin view sits behind a multi-factor sign-in. A key can be replaced or deleted there, but never viewed.
• Error logs record the call and its status code, never your key or your ticket text.

Our privacy policy lists every provider we use.

How to revoke access

• Switch off the Cobotic user, or set its API application to inactive, in your PSA. Cobotic stops at its next call.
• Switch off any single AI workflow from your Cobotic dashboard.
• For the Test Drive, delete the application or set it to inactive the moment the run finishes.

Questions or a security concern

Ask us anything before you give us access to anything. Use our contact form and we will answer your security questions first.