AI customer support agents that triage questions and draft answers for your okay
Support doesn't have to mean a full inbox. Brohns builds a team that sorts incoming questions by urgency and topic, and drafts clear answers grounded in your own knowledge — the tricky ones flagged for you, nothing sent without your approval.
A focused agent for each job.
Sorter · Triage
Sorts incoming questions by topic and urgency.
Responder · Support
Drafts a helpful reply from your knowledge, for your approval.
From a sentence to a working team.
Connect your inbox
Connect Gmail and new mail arrives as tickets — no DNS or webhook setup. On another mail host, forward it to the inbound endpoint instead.
Triage automatically
Every ticket is classified in one pass — category, sentiment and priority — so nothing urgent slips through.
Draft the answers
The Responder writes replies grounded in your policies and reusable answer templates; sensitive ones wait for you.
Approve and send
You review the drafts and send — the agent handles the rest.
Only what this goal needs.
- Email (Resend or SMTP)
- Optional: WhatsApp/SMS inbound via Twilio — a text becomes a ticket, the reply goes back on the same channel
- Optional: your store policies and answer templates as the grounding source
- Optional: a public FAQ widget for your own site — answers grounded in your policies, honest when it doesn't know
You never connect an AI key — Bro and the agents are built in. Keys live in an encrypted vault, read server-side only for the action you approved.
Powerful agents. You stay in charge.
Faster first replies
Draft answers are ready the moment a question lands.
SLA timers, honest ones
Each ticket gets a first-response deadline from its priority; at-risk and breached tickets surface first.
A team inbox
Assign tickets to workspace members and filter by owner — nothing urgent waits on 'whose is this?'.
Nothing urgent missed
Triage surfaces what needs you first.
Your voice, your call
Every reply is a draft until you approve it.
Good to know.
Will it reply to customers automatically?
No — replies are drafted for your approval; you decide when an agent may send on its own.
Can it ever answer on its own?
Only if you opt in, and only the safe categories — order-status and product questions. Money, complaints, cancellations and address changes always wait for a human; a hard daily cap and injection checks are enforced server-side.
Where do answers come from?
From your shop policies and templates — replies never invent a delivery date, return window or refund amount; missing policy means the agent honestly says it will check.
Can it flag sensitive messages?
Yes — the tricky ones are held for you rather than answered automatically.
Whose customers does it recognise?
Email triage matches an incoming message to someone already in your account — a customer with an order, or a contact you reached out to. A question from a complete stranger has nothing to ground an answer in, which is what the public FAQ widget is for: it answers from your policies and says honestly when it doesn't know.
Ready to put a team on support?
Free to start. No credit card. Nothing goes out without your approval.