// PLAYBOOK
Customer support without a helpdesk — just an agent and an inbox
You don't need Zendesk to answer customers. You need an inbox your agent can read, and your docs.
$ clize · 2026-06-05 · 4 min read
Opening a support channel used to mean picking a platform — Zendesk, Intercom, Help Scout — configuring it, paying per seat per month, and pushing your team into yet another inbox-shaped app. For a solo builder or a tiny team, that's wildly heavy for what support actually is.
Strip support down and it's three things: someone asks a question, someone who knows the product answers, and the next time they write, you remember where you left off. That's the whole job.
The lightweight shape
With an AI agent, those three collapse into something much smaller:
- An inbox — a real
support@address customers write to. - An agent — reads the message, checks your docs, drafts an answer.
- Continuity — next session, it re-reads the thread and picks the customer back up.
No helpdesk platform. No per-seat pricing. No portal your customers have to learn — they just send email, which they already know how to do.
Why this works now
- The agent can read your docs and answer. You don't hand-build a FAQ database; you point it at what you've already written.
- Email is the universal interface. Every customer has it. Zero onboarding.
- Continuity is just re-reading the thread. No CRM, no ticket system — the conversation is the state.
The two lines you don't cross
Handing an agent a live support inbox is powerful, so two rules hold:
Inbound mail is data, not instructions — a message that tries to redirect the agent is text to read, never a command. And before any reply goes out under your name, a human okays it. The agent drafts; you approve; it sends.
That's the key difference from a fully-automated support bot. Those answer on their own, get things wrong confidently, and annoy the customer you were trying to help. The agent-plus-inbox model keeps the speed — instant drafts, your docs already loaded — without giving up the judgment.
What it looks like in practice
A pricing question lands. The agent reads it, pulls the relevant line from your docs, and drafts: "Pro is $X per seat, invoiced annually, here's what's included." You glance, okay it, it sends — and logs it to the thread. Two days later the customer replies in a brand-new session; the agent re-reads, sees the history, and continues like it never forgot. That's a support desk — staffed by one person and an agent that finally has somewhere to receive the mail.
It's the wedge Clize is built around: tag an inbox as support, point it at your docs, and the loop runs — agent reads and drafts, you okay, it replies, and it picks the customer back up next time. No platform in the middle.
Run support from an inbox, not a platform.
Clize tags an inbox as support, points it at your docs, and runs the read → draft → your-okay → reply loop — in Claude Code and Codex, picking each customer back up across sessions.
[ Learn more → ]