— Door 02 · Show, don't tell
The onboarding playbook, already started.
Your brief says "build the playbook as you go." So I started it. This is a real onboarding playbook for getting a typical Cliniko allied-health clinic live and confident on Aeva: a 14-day onboarding journey, a go-live setup checklist, a plain-English script library that explains AI to a 65-year-old without making them feel stupid, real AI prompt work for an Aeva-style call flow, and a simple health/risk model for calling clinics before they fall off. It's concept work for the application — a generic clinic, not a real Aeva customer — but it's the actual artefact, not a description of one.
— 01 · The first 14 days
Signup to confident go-live, mapped
The journey every clinic runs through — from the moment they sign up to the day they trust Aeva with their phones. Teal = AI/automation does the heavy lifting; coral = the human moments that earn the relationship and prevent churn.
Welcome + instant orientation
Automated welcome the moment they sign up: what happens next, the one thing to prepare (Cliniko access), and a booked setup call. No clinic ever sits wondering "now what?"
AI / automated
The setup & connection call
A live video call: connect Aeva to Cliniko, confirm the clinic's hours, services and booking rules, and listen to how their reception actually talks. This is where I'm strongest — patient, jargon-free, doing the technical setup with them, not at them.
Human · the go-to person
Script & prompt configuration
Tune Aeva's call flow to the clinic — greeting, services, the booking/reschedule/cancel logic into Cliniko, the edge cases ("we don't do that here," after-hours, urgent triage hand-off). AI-assisted drafting; my prompt work makes it sound like them. (See section 03.)
AI-assisted · my prompt work
Test calls & the Cliniko write-back check
Run real test calls together: book, reschedule and cancel an appointment and watch it land correctly in Cliniko in real time. Nothing goes live until the clinic has seen it work with their own eyes.
Verify · together
Go-live + the confidence handover
Forward the phones (or after-hours first, if they're nervous), and walk them through what they'll see: where bookings appear, how to listen back, who to call if anything feels off. The goal isn't "switched on" — it's "switched on and not anxious."
Human · get them confident
Week-one confidence call
Proactive call, not a wait-for-a-ticket: "How did the first week feel? Any calls that went sideways?" Fix the one prompt that's bugging them. Week one is where trust is won or quietly lost.
Human · proactive
Activation review — are they getting value?
Check the activation signals (calls answered, bookings made in Cliniko, after-hours catches) and show the clinic the value in their own numbers. If the signals are weak, this is the "call before they fall off" moment — see section 05.
Human · retention trigger
— 02 · Go-live setup checklist
The checklist a clinic is never live without
The repeatable checklist behind the setup call — so onboarding is consistent across every clinic and nothing is missed. This is the kind of artefact "build the playbook as you go" produces.
① Connect & confirm
- Cliniko connected and authorised; correct business + practitioner calendars selected.
- Opening hours and public holidays confirmed (drives after-hours behaviour).
- Services & appointment types mapped to Cliniko exactly (names, durations).
- Booking rules confirmed: new vs returning patients, double-booking, buffers.
- Phone number / forwarding path decided (full, overflow, or after-hours first).
② Voice & edge cases
- Greeting in the clinic's own words; clinic name pronounced correctly.
- Reschedule & cancel flows tested end-to-end into Cliniko.
- "We don't do that" responses for services they don't offer.
- Urgent / clinical triage hand-off rule agreed (when to take a message or escalate).
- Test calls passed and the clinic has watched a booking write back live.
— 03 · Real AI prompt work
An Aeva-style call flow, actually written
This is the part of the role that's my core craft. Below is genuine prompt work I wrote for an AI receptionist call flow — the kind I'd configure per clinic. Not a screenshot of someone else's; written here, for this application.
system_prompt · clinic_reception_agent · v1
Written by Khalid
# ROLE
You are the phone receptionist for {{clinic_name}}, an allied-health clinic.
You are warm, calm and efficient. You never rush. You never sound like a robot.
Most callers are patients booking, moving or cancelling an appointment — some are
older or anxious. Speak in plain language. Never use clinical or technical jargon.
# WHAT YOU CAN DO (via Cliniko)
- Book, reschedule and cancel appointments in real time.
- Match returning patients by name + date of birth before changing anything.
- Offer the next available times for the requested {{service}} and practitioner.
# GUARDRAILS
- You do NOT give clinical advice. If a caller describes a medical emergency,
calmly tell them to call 000, then end warmly.
- If asked for a service the clinic doesn't offer ({{excluded_services}}),
say so kindly and, if known, suggest they speak to their GP.
- If unsure or the caller is distressed, take a message and assure them a human
will call back — never guess.
# STYLE
- Confirm every booking back to the caller in full ("So that's [day], [time],
with [practitioner] — have I got that right?").
- One question at a time. Short sentences. Let them finish.
Why this is the right work: good AI-receptionist onboarding isn't just flipping a switch — it's getting the prompt and guardrails right for each clinic so the AI books correctly, stays inside its lane, and sounds human to a nervous patient. Tuning that per clinic, then explaining it back to the owner in plain English, is exactly the AI-prompt-plus-translation work this role asks for — and exactly what I do every day.
— 04 · Plain-English script library
Explaining AI to a 65-year-old, without the jargon
Your brief tests this directly. Here's how I'd actually answer the questions clinic owners ask — replacing the jargon answer (grey) with the human one (teal). This is a real on-call skill, not a personality trait.
"So... is a robot answering my phone?"
✗ "It's an AI voice agent using natural language processing to handle inbound telephony."
✓ "Think of it as a really reliable receptionist who never gets sick, never goes to lunch, and never misses a call — even at 9pm. It answers, books people in, and everything lands in your Cliniko just like your front desk does it. You're always in control."
"What if it gets something wrong?"
✗ "The model has guardrails and fallback logic with escalation pathways."
✓ "Good question. We set it up so that if it's ever unsure, it takes a message and tells the person you'll call back — it never guesses. And you can listen to any call. We'll test it together before a single real patient hears it."
"I'm not techy. Is this going to be hard?"
✗ "You'll need to configure the integration and provision the call routing."
✓ "You don't have to do any of the technical bits — that's my job, and I'll do it with you on a call. Your part is just telling me how your clinic likes to do things. If you can describe a normal day at reception, you can set this up."
"Will my patients know it's not a person?"
✗ "It's a conversational TTS system optimised for naturalness."
✓ "It speaks naturally and warmly, and it uses your clinic's own greeting and words. Most callers just feel like they got through quickly and got booked in — which, honestly, is the bit they care about."
— 05 · Call them before they fall off
A simple health model that flags risk early
"Call them before they fall off" is the highest-value line in the brief — but it only works if you know who to call. Here's a simple, no-fancy-tooling-required health read I'd run from week one. (Built out further in Door 03.)
● Healthy · leave them be
Activated & humming
- Calls answered & bookings landing in Cliniko
- After-hours calls being caught
- Logged in / listened back recently
- No "turn it off" or confusion signals
Action: light touch — a value recap at Day 14 & Day 30, then leave them to it.
● At risk · reach out now
Quiet or hesitant
- Low call volume / forwarding not fully on
- Hasn't logged in since go-live
- One bad call they mentioned but didn't escalate
- Went silent after an eager start
Action: proactive call this week — "noticed it's been quiet, can I help tune it?" Fix the one thing.
● Churn signal · call today
Slipping away
- Forwarding turned off / phones back to front desk
- A complaint or "it's not working for us"
- No bookings written to Cliniko in days
- Ignored the last two check-ins
Action: personal call today, founder looped if needed — understand, fix or save before they cancel.
Honest note: Everything here is concept work produced for my application — a real 14-day journey, a real checklist, real prompt work and a real health model — built for a generic Cliniko allied-health clinic, not a real Aeva customer, and not Aeva's official onboarding process. The specifics of Aeva's product, Cliniko's exact integration behaviour and your real onboarding metrics are things I'd learn in the first weeks and refine this against. The prompt is illustrative of how I'd configure a call flow, not a claim about Aeva's internal system. See Door 01 for the honest fit map.