Everything the agent touches.

01 - Autopilot

The AI agent answers, and shows its work.

• Every reply starts from your help centre, your policies, and your product data - not from what the model happens to remember.

• Before anything is sent, a verification pass checks every claim against the source it came from - a reply that cannot back itself up does not go out.

• When confidence drops, your agent stops and hands the conversation to a human - with the whole trace attached.

02 - Inbox

One console your team actually runs on.

• Every channel in one queue, with the customer's orders and history beside the thread.

• A copilot drafts the reply, cites where it came from, and waits for a person to press send.

• When your team keeps correcting the same thing, that becomes a proposal - approve it once and the agent answers that way from then on.

03 - Flows

Complex inquiries, finished end to end.

• Multi-step conversations for returns, exchanges, and troubleshooting - written in plain language.

• Describe the journey you want - or have the assistant draft it - then review every step and test-run the flow before it goes live.

• The agent still answers whatever else the customer asks along the way.

• When a customer sounds like a buyer, a flow can capture the details and hand the thread to your team with everything attached.

04 - Channels

The same agent, wherever they write.

• Your customers write where they already are; your agent answers there with the same policies and the same customer record.

• Keep your addresses and handles; connect a channel in a few clicks and switch it on when you're ready.

05 - Analytics

Read what your customers keep asking.

• Resolution rate over time, split by what the agent finished and what it handed over.

• Topics clustered from the conversations themselves - nobody has to tag anything.

• Trends you can act on: a spike in delivery delays is a warehouse problem, not a support problem.

06 - Testing

Try it on your worst tickets first.

• A sandbox that runs the real agent against real questions - nothing reaches a customer.

• Every test turn opens up: what the agent knew, what it found, what it did, and why - in plain language, not logs.

• Go live channel by channel, in draft mode for as long as you like.

07 - Learning

It learns where it actually struggled.

• Your agent watches its own closes: what it resolved, what it handed over, what your team had to fix.

• Gaps become drafted articles and recurring edits become writing rules - each one a proposal with approve and undo, never a silent change.

• You decide how much it may apply on its own, domain by domain.

08 - Assistant

Run the whole thing in plain language.

• Describe what you want in a sentence; the assistant makes the change for you, from a single tag to a reworked automation.

• There is no settings maze to learn. If you can say it, you can usually ship it.

• It knows your whole setup end to end - and what it can do for you grows every month.

09 - Integrations

Connected to the systems that hold the answer.

• Connect your store once, and your agent reads orders, fulfilment, and refunds from the source.

• Moving from another helpdesk? Bring your tickets and help centre with you, so day one is not a blank slate.

• Anything else through your own APIs, with the actions you allow and nothing more.

10 - Security

Six facts, no badges.

This is the whole list. The certificates behind the hosting facts are held by our infrastructure provider and are available on request.