Start with one role. Expand when it is working
AI staff take repetitive work: support answers, lead follow-up, operations organizing — start with the role that eats the most hours. It plugs into the website, WeChat and email you already use; no system swap. Price and commitment questions always go to a person; it never improvises promises. Before it faces customers, it is tested on your real past questions, and it only opens if it passes.
Who it fits, and what gets in the way
Support inboxes that are the same dozen questions on repeat
Teams with plenty of leads and follow-up riding on each person's memory
Shops unstaffed at night and on weekends while enquiries keep coming
Support answers the same dozen questions all day, while the genuinely hard ones wait in the same queue
Leads come in, follow-up rides on memory, and the ones that slip away slip silently
Nights and weekends are unstaffed; a question asked at 9pm is answered after the customer has booked elsewhere
Self-serve AI tools were tried: the output still needed line-by-line proofreading — an intern you have to supervise
What you get
✓ A role trained on your workflow and your tone: support, follow-up or organizing — pick one first
✓ Plugged into your existing website, WeChat and email
✓ Tested on real historical questions before launch; live only if it passes
✓ Managed operation is included after launch: daily monitoring, corrections and content updates — not left on your desk
✓ Tuned monthly, with a report: how much it answered, how much it handed to a person
✓ A quarterly results review against the metrics set at launch: volume answered, hand-off rate, where it went wrong
| Component | What is included | Deliverable |
|---|---|---|
| Role setup | Trained on your workflow and your tone: support, sales follow-up, operations organizing — one role first | A role brief: what it takes, what it never takes, in writing |
| Channel hookup | Runs on your existing website, WeChat and email; no system swap | A live role on your own channels |
| Pre-launch testing | Tested on your real historical questions; opens to customers only if it passes | Test record and pass verdict |
| Managed operation | Daily monitoring, error correction and content updates handled by us — no line-by-line proofreading on your side | Monthly report: answered, handed off, changed |
| Quarterly review | Measured against the metrics agreed at launch; expand or adjust, on the numbers | Quarterly review report |
Tiers
Pilot Role
- One role first, on the channels you already use
- Live in 2–3 weeks
- Expand only when it passes; if it falls short, we tune first
Multi-Role Managed
- Several roles working together, hand-offs logged and reviewable
- Monthly tuning, plus a quarterly results review reported against the agreed metrics
- A dedicated dashboard showing what each role did
How it runs
Pick the role, gather material
The question type that repeats most, prepped from your past chats and price materials
Train and test privately
Trained on your workflow and wording, run internally first, never straight at customers
Pass the test, then open
A round of your real questions first; price and commitment questions always route to a person
Tune monthly
One report: how much it answered, how much it handed over, where it went wrong
Start with one role. Expand when it is working
Tell us where it hurts first. Scoping and a written quote come before anything starts; looking costs nothing.
Or ask Leo, the assistant at the bottom right, first — get the questions clear before you write to us.
Two newer roles: answering calls, catching missed ones
AI voice reception takes inbound calls only: hours, address and services answered on the spot, booking details taken and passed to you. Usage runs on your own voice-service account; we handle training, launch and monthly tuning, and never touch your billing.
Missed-call text-back fires an automatic text after a missed call, picks the conversation up and hands it over once the need is clear. Texts are billed per message at your cost, with the rate written down before anything is switched on. Both roles keep the same line: anything they cannot answer goes to a person.
Boundaries
Not included
- It never takes pricing or commitment questions: those always route to a person, by rule
- It does not replace your systems: it runs on your current channels, and your current staff keep the work that matters
- It is not promised error-free: what is promised is the spot-check, correction and hand-off machinery, written into the service terms
What we need from you
- Past Q&A and pricing material, for prep and testing — the more real, the better the test
- A hand-off target: which questions go to whom, agreed before launch
- Launch metrics: the few numbers we will judge it by, agreed together
How engagement and quoting work
Start with a pilot: the question type that repeats most, fixed scope, fixed timeline, live only after it passes testing on your real questions. Roles are priced per role by a person after discovery, written down before work starts. Managed operation is included after launch, tuned monthly and reviewed quarterly. Add roles per role — no need to buy the whole roster up front.
Questions on this page
What if it answers wrongly?
Boundaries are set first: pricing and commitment questions always route to a person. It is tested on your real questions before launch, and wrong conversations are reviewed and corrected one by one each month.
Do we have to replace our current support staff?
No. It takes the repetitive questions; people take the hard and the important ones. The split is written into the plan.
How is this different from buying an AI tool subscription?
A tool hands you the work: configuration, proofreading, corrections. This is a managed service: we configure the role, spot-check its conversations, correct its errors and report monthly. You read results, not transcripts.
Will customers know they are talking to an AI?
Yes. The role introduces itself as an AI assistant and hands off what it cannot answer. It never poses as a person.
How do we start?
Describe the question type that repeats most, in a few sentences on the contact page. We scope the pilot role — range, timeline, fixed price in writing — before anything starts.
Related
Start with one role. Expand when it is working
Tell us where it hurts first. Scoping and a written quote come before anything starts; looking costs nothing.
Or ask Leo, the assistant at the bottom right, first — get the questions clear before you write to us.
