UX ResearchRe-designWebsite
Caffe Yoga
A self-initiated redesign helping visitors explore classes, instructors, memberships and events.
View caseConversational UXConversation FlowAssistant
Taskora Booking Assistant
From problem to booked slot
A full scope from problem to booking chat bot: diagnose → match trade → book slot.
Product Overview
A conversational booking assistant that lets someone describe a household problem in plain language and reach a confirmed appointment, without knowing which trade they need or reading a service list.
The assistant asks a short diagnostic sequence, maps the answers to the right professional, then hands over an upfront price and the next available slots.
- Role
- Conversation Designer
- Timeline
- 6 weeks
- Scope
- End-to-end flow
The problem
People know what is broken, not who fixes it.
- Category-first booking asks people to self-diagnose before they can start, which is exactly the step they are unsure about.
- A conversation can gather the same information in the order people naturally give it, and confirm the match back to them.
Process
01
Intent Mapping
Real requests were collected and grouped into the intents the assistant has to recognise.
02
Conversation Flow
A happy path plus repair paths were drawn, so the bot can recover from vague or partial answers.
03
Persona & Tone
A calm, practical voice was defined with sample turns for confirmation, error and hand-off.
04
Prototype & Test
The script was prototyped and tested in wizard-of-oz sessions to catch dead ends.
Intents
Collected requests were grouped into four intents the assistant has to recognise from the very first message.
Describe a problem
“Water is dripping under my kitchen sink.” The most common entry point, and never a trade name.
Ask for a trade
“I need an electrician.” Already diagnosed, so the assistant skips straight to scoping the job.
Ask about price
“How much does a boiler service cost?” Answered with a range before any booking pressure.
Manage a booking
“Can I move Thursday to the weekend?” Reschedule and cancel handled in the same thread.
Conversation flow
The happy path runs down the middle, with a written repair path for every way the conversation can slip: a vague answer, a wrong diagnosis, a request outside scope, no availability, or silence.
Sample dialogue
The happy path written out, from an undiagnosed symptom to a confirmed booking.
Hi, I'm the Taskora assistant. What's going on at home?
There's water under my kitchen sink.
Got it. Is the water dripping slowly, or is it running while you use the tap?
Only when I use the tap.
That sounds like a leaking waste pipe, so a plumber is the right fit. Shall I look for one near you?
Yes please.
Plumbers near you start at €65 for a call-out. Earliest slots are today 16:00 or tomorrow 09:00.
Tomorrow morning.
Booking a plumber for tomorrow at 09:00, leaking waste pipe under the kitchen sink. Confirm?
Design principles
01
One question per turn
Each turn asks for a single piece of information, so nothing has to be re-read or re-typed.
02
Confirm before you commit
The trade, the job and the slot are each read back in plain words before anything is booked.
03
Every dead end has an exit
Vague answers, wrong guesses, no availability and silence all have a written recovery path.
04
Human hand-off on demand
Asking for a person is treated as a valid intent at any point, not as a failure state.
Outcome
A written conversation that carries someone from a described symptom to a confirmed appointment in nine turns, with a repair path for every point where the conversation can break.
What I learned
01
People describe symptoms, not trades, so diagnosis has to come before matching.
02
Every question the bot asks must earn its place, each one is a chance to drop out.
03
A clear hand-off to a human is a feature, not an admission of failure.
Explore more work
Available for new projects
Interested in working together?
Whether you're building a product, improving an existing experience, or looking for thoughtful UX feedback, I'd love to collaborate.