What does "on-site pairing" actually look like?
Our team pairs with you for two weeks — in person when you're in Auckland, on Discord / Zoom / WeChat video everywhere else. We shadow a real day of your work, design the roster with you in the same Slack you already use, and ship tuning changes live. After the intensive, we stay on monthly to watch, tune, and promote agents.
What's "the ladder"? Do agents actually get promoted?
Yes. Every teammate starts as an intern (borrows your identity, drafts everything for your ✓). When they've earned trust they graduate to a shared teammate your crew can @mention, then an employee with their own email and signature, and finally a customer-facing colleague. Each promotion is a decision we make together — we don't auto-upgrade anything.
When an agent faces my customers, is the AI disclosed?
Always. Any teammate at the external scope carries an "AI teammate · disclosed" badge in its signature and replies. It also offers a human handoff on every message. We help you draft the disclosure wording so it fits your brand — but we don't do ghost-AI-pretending-to-be-human. That's a hard line.
Why Discord? I don't use it.
Discord is just our default — the agents aren't tied to it. Same posse, same memory, deployed wherever your customers and team already live: WhatsApp, 微信 / 企业微信 (WeChat / Work WeChat), Slack, Microsoft Teams, Telegram, Instagram DM, SMS, email, a web chat widget on your site, or a custom dashboard we build for you. Pick the one channel your customers use — or run the posse across several at once. We default to Discord only when there's no preference, because it's free, cross-platform, and handles multi-agent threads cleanly.
Are these just ChatGPT with prompts?
No. Each agent has its own tools (gmail API, calendar API, your Notion, your sheet), its own memory, and explicit limits. They can actually do things, not just talk about them.
What if an agent does something wrong?
Anything with side effects (sending, booking, paying) requires your ✓ by default. Our team relaxes this per-agent only after track record, and only at the scope where it's appropriate.
Who sees my data?
Only you + our small build team. Models run on infra you can audit. We never train on your data, full stop. You can self-host the whole posse on your own hardware if you prefer — configs are yours to walk away with.
Which languages work?
Any language your customer writes in, the agent replies in. English, 中文, 日本語, 한국어, हिन्दी, Deutsch, Français, Español, Tagalog, te reo Māori — detected per-message, no configuration on your end. The same agent handles a Mandarin enquiry at 2am and an English one at 9am without you switching anything. For low-resource or dialect-heavy languages we run a short pilot first so you see the quality before committing.
Can I add / retire / promote agents later?
Yes. Posses are meant to evolve. Each new agent is a small fixed build fee; retiring one is free; promotions up the ladder are part of the monthly care cycle.