hospitable

Head of Design case study

Managing listingsacross channels.

The case: how hosts keep one listing right on every channel today, what I found, and how I’d approach it and test it.

Antonio Silva

Introducing

Smart Listings

Set it once, earn more.On every channel.

A new way to run every channel: you stay in control, work faster, and keep improving.

IntroWhy it matters

Keeping a listing good on every channel is a part-time job

Every channel has its own guests, competition and rules. Hospitable manages the listing as one thing, in one crowded place.

Set once, the listing drives revenue.
AI can keep tuning it.

Guests look for

Vibe and photos
Facts and amenities
Space for families

Every kind of host

P1Family hostWants it to just work

P4New to a PMSNot sure what it’s for

P2SEO optimiserTunes channels weekly

P3Portfolio managerNeeds it at scale

Hospitable today

No view per channel Listing spread across screens Field-by-field edits

There is no one size fits all for channels. Guests behave differently on each one. They have different goals, different references and different competition. What guests look for on each channel is my assumption, not data yet, and the card says so. What is a fact: Booking.com writes its own description. So each channel needs its own strategy. [Press] And every kind of host matters. The family host just wants it to work. The newcomer is busy with everything else. They aren’t sure what the listing tools are for, or that the listing could be better at all. The optimiser tunes every channel every week. And the portfolio manager needs it at scale. One product has to serve all of them. [Press] Today there’s no view per channel. The listing is spread across screens, and tailoring a channel means editing field by field. [Press] And that matters. Once it’s set, the listing is what drives revenue. AI can keep tuning it. About 50 seconds.

IntroThe brief

The brief asked for an intelligent listing. I took it deeper.

The brief asked

“If we remove the constraint of ‘one field to rule them all’… what does that product look like? … What level of control and transparency is needed?”

My spin

I took the brief's question deeper:

From copy to strategy. A view and a goal for every channel, not just words.
From the text to the whole listing. Tips for pricing, availability and policies too, built on data.
You choose the autonomy. Review every tip, or turn on an optional Autopilot inside your guardrails.
From launch to results. Tracked over time, before and as it evolves.

I started from two questions. What does a listing look like when each channel can be different? And how much control do hosts need? Then I went deeper, in four ways. [Press] First, from copy to strategy. Every channel gets its own view and its own goal, not just new words. Second, from the text to the whole listing. The tips cover pricing, availability and booking policies too. They’re built on data: guests’ questions, the facts about the home, and similar homes. Third, and this matters most, you choose how much freedom it has. You can review every tip yourself. Or you can switch on an optional Autopilot that works inside your guardrails. You decide how much you stay in the loop. Fourth, from launch to results. Every strategy is tracked over time, against the weeks before it and week by week as it changes. So hosts keep what works. About 40 seconds.

My hypothesis

If every channel gets its own strategy, and recommendations backed by data, listing management stops being a chore and becomes Hospitable's competitive advantage.

It's true when

Changes take less time Hosts trust what goes out Hosts keep the tips More bookings on every channel

IntroThe idea

Listings that manage themselves.
Hosts who stay in charge.

One base

Facts about the home, written once. Each channel reads what it needs.

A strategy per channel

Every channel has its own guests and competition. Set a goal, an audience and guardrails for each one.

Tips backed by data

Listing, pricing, availability and policies, each tip with its reason: guest questions, home facts, similar homes.

You stay in control

Review every tip, or turn on Autopilot inside your guardrails, and choose what it may touch. Money always asks first.

Measured

Each strategy tracked over time, from the weeks before it onwards.

One field, four cases

IntroThe shift

From managing fields to managing outcomes

Today, a host edits field by field and hopes it works.
With Smart Listings, they set a goal and AI helps them reach it.

Today With Smart Listings The listing One editor for everything, in one place One Base, and a view per channel Strategy One copy, overrides scattered A goal and an audience for each channel Recommendations Improving it is guesswork AI in every tip, for listing, pricing and policies, with its reason Control Every change is manual Review every tip, or turn on an optional Autopilot Context No before, no after Tracked over time, before and as it evolves

Product audit

IntroGoals

What success looks like for Smart Listings

Five goals, each with the metric I'd track and where I'd expect it 30 and 90 days after launch.

GoalWhat we measureToday30 days90 days

More bookings and requestsBookings and booking requests per listing, against its own weeks before, from the bookings Hospitable already hasBaseline+5%+7%

Hosts adopt itEligible properties with a strategy on at least one channelNew10%25%

Hosts keep refiningAdopters who act on a tip in the monthNew60%50%

Tips earn their placeTips accepted, out of tips shownNew30%40%

Fewer losses, less supportListing and sync tickets per 100 propertiesBaseline−5%−10%

How I set them

IntroProcessWith AI

The process I'd runHow I'd agentify it

I ran what I could reach, sampled what needed hosts, and planned the rest.Agents read and make at every stage. At every gate, I decide.

Done

Two parts. First, the process I’d run. These are the six stages, with the methods under each. Done is what I did here. The product audit and heuristic evaluation. Desk research on channels and competitors. Seven directions on a weighted matrix. The prototype in code, reviewed round after round. The design system. And the goals, roadmap and kill criteria, written out. Sampled is a labelled stand-in. I had no access to hosts. So I sampled five interviews, a card sort and the support data, to see what the main issues might be. The concept rests on those, and on what I could see for real. Planned is what I’d run with real access: every test of the design, the pilot and the rollout. Second, how I’d run it with AI agents. At each stage, something goes in. One of my skills does the reading or the making. Then there’s a gate where I decide. Which hypotheses matter. Whether it’s a pattern or a loud few. Which direction. Whether the system holds. What a finding means. And what ships. The agents carry the work. The decisions stay mine. About a minute.

DiscoveryProperty management, mapped

Today, the listing isn't a place in Hospitable

I mapped property management end to end, in a real trial account.

Everything hangs off the property
Ten sections, in menu order
Each listing is a row, not a page
Channel settings live inside sections

Property10 sections

Overview Airbnb Booking.com Direct

DetailsPer channel

PricingPer channel

AvailabilityPer channel

Calendar

Knowledge Hub

Booking policiesPer channel

Guest experiencePer channel

Operations

Direct

SoEach channel's listing gets its own pages, with tips on every channel.

A real trial account with a real Airbnb listingHospitable's help centreThe full map

I started by mapping property management end to end. I used a real trial account, with a real Airbnb listing in Berlin. Not the whole product, just the part where a host sets up and runs a property. Everything hangs off the property: ten sections, in the product’s own menu order, from Overview to Direct.

[Press] The listing, the thing a guest reads, isn’t one of those sections. Each connected channel, like Airbnb, Booking.com or Direct, is a row on Overview. None gets a page of its own.

[Press] What can differ per channel lives inside the sections. Details, Pricing, Availability, Booking policies and Guest experience each have their own channel settings, field by field.

[Press] So each channel’s listing gets its own pages, with tips on every channel. The full map, every section with its pages, and how I made it, are in the detail. About 40 seconds.

DiscoveryProduct audit

The Knowledge Hub repeats the listing, and has no job of its own

Its copy drifts, the AI answers from it, and guests’ questions go unseen.

Details

Amenities

WasherAll channels

HeatingAll channels

Knowledge Hub

Amenities

Washing machine

Sources (1)

September 29, 2026

Heating

Sources (1)

September 29, 2026

Wheelchair accessible

Sources (1)

September 29, 2026

Ask the AI

Is the apartment wheelchair accessible?

Yes, the apartment is wheelchair accessible!

On a trial listing

Same facts, written twiceSource-of-truth claritySev 4

Edits to Details don’t reach it

Answers from the copy, not the listingAI transparencySev 4

SoEach fact lives once, in the Listing base. The Hub reads it and learns from guests’ questions.

Recreated from a trial listingHospitable’s help centreSeverity 0 to 4, one evaluatorHow the Hub works today

This is the slide the case rests on. So I made sure it’s true for any host, not just my trial account. Next to every listing, Hospitable keeps a Knowledge Hub. Today it repeats the listing. The amenities, the rules and check-in are written again as separate entries, each with one source and a date. In my heuristic evaluation, that breaks source-of-truth clarity, at severity 4.

[Press] The Hub is filled once, when the host presses Generate, from Details and past messages. Hospitable’s help centre says it reads Details only once. So later edits never reach it. It drifts by design, not by accident.

[Press] And the guest AI answers from that copy, not from the listing. On my trial listing, Ask the AI told a guest the apartment is wheelchair accessible. No amenity says so, and no source is shown. That breaks AI transparency, also a 4.

[Press] Meanwhile, the questions guests really ask stay in the Inbox, and nothing learns from them. So the Hub has no job of its own. My answer: every fact lives once, in the Listing base. The Hub reads it, and its job becomes guests’ questions. The severities are my own judgement, so I mark them as assumptions. About 60 seconds.

DiscoveryHeuristic evaluation

No screen shows one listing on one channel

Hospitable keeps per-channel values field by field, never as one listing.

What does my Airbnb listing say?

Details Pricing Availability Booking policies Direct

~15fields

No page shows the answer

Per-channel awarenessSev 4

7 ways to say “different on each channel”Consistency and standardsSev 3

€25per stay1 channelsEdit per channelFees
1 channelWi-Fi
ShampooAll channelsAmenities
Airbnb deposit€0Deposits
%Markups
AirbnbCancellation
ChannelAccept allLow riskReviewAirbnbInstant bookThe one channel column

SoEach field saves on its own: a change reaches the channels that read it. A custom value stays on its channel.

Product auditSeverity 0 to 4, one evaluator

The second gap in the structure is the channel view. Say a host asks the simplest question: what does my Airbnb listing say right now?

[Press] To answer it, they hop across five sections: Details, Pricing, Availability, Booking policies and Direct. They check about fifteen fields, one info icon at a time. And at the end, no page shows the answer. In the heuristic evaluation, that breaks per-channel awareness, at severity 4.

[Press] Hospitable can hold different values per channel, but only field by field. And it shows that in seven different ways: an Edit per channel link, a scope caption on its own, an All channels label, a field that belongs to one channel, a row per listing, a section per channel, and a channel column. That breaks consistency and standards, a 3. There’s also a trap. Opening Edit per channel just to look marks the form as changed. So a host who only looked, then saved, has created overrides.

[Press] Exactly one screen, Instant book, treats the channel as a column. So the product already knows the pattern, and it works for every channel.

[Press] So in Smart Listings, each channel gets its own pages. Every field there is one of four cases. It uses the Base. It’s custom for that channel. It exists only on that channel. Or the channel sets it itself, like the description Booking.com writes. That one is read-only, with a lock that says why. Looking never changes anything. About 50 seconds.

DiscoveryData analysis

I want to see where listing trouble costs hosts and deals

Not run yet.

Field overrides
1Which fields do hosts override most, and is any text field among them?
Support ticketsKnowledge Hub
2How often does a ticket say the assistant contradicted the listing?
Support tickets
3How many tickets ask where to edit something, and what do they cost to resolve?
Sales calls
4Does listing trouble rise with portfolio size in sales calls, and when does it become a reason not to buy?
Public feedback board
5What do hosts ask for on the board without being asked?

Planned pulls: 12 months of tickets (300 tagged, 100 untagged), 618 Gong calls with 60 listened to, the field-override audit, Knowledge Hub usage, and the public board.

SoIt sets what tips must cover.

618 calls, 457 accountsPulls not run, no Hospitable data yet

DiscoveryResearch questions & assumptions

I re-ranked my riskiest bets when evidence moved

The hypotheses behind a view per channel, scored by importance against evidence. Booking.com's own docs partly refuted one.

Refuted, then turned into a rule

Writing back to Booking.com was my number two bet. Booking.com writes its own description, so a host's voice there is the quality of their facts.

Test these first

1Trust in AI tips
2Who wants a voice per channel
3What Autopilot may touch

Booking.com and Hospitable help pages

I turned my thesis into hypotheses. Each one says what would prove it wrong, and I scored them by importance and evidence. These are the thirteen that decide Smart Listings. Write-back to Booking.com ranked second. Then Hospitable’s help centre and Booking.com’s partner docs showed that Booking.com writes the description itself.

[Press] So it left the list and became a design rule. On Booking.com, a host’s voice is the quality of their facts.

[Press] What I test first: does a why, a source and an undo make hosts trust AI tips? Who wants a different voice on each channel? And what will hosts let Autopilot touch? Almost nobody asks for AI listing copy in public, so that demand is a bet too. Three of my first five riskiest bets were about import. They wait for the next release. About 50 seconds.

DiscoveryResearch plan

Data first, then people, with an owner for every unknown

Ten methods over four weeks. Hosts can be simulated to show the method. What Hospitable's own teams know never is.

Prep

Get ready

Which assumptions would sink the design?

Kick-offAssumption mappingData accessRecruiting

Week 1

Data and teams

What can Hospitable write on each channel, and how fast?

Sales calls (Gong)Ticket analysisField override usageKnowledge Hub usageStakeholder interviews

Weeks 2 and 3

Hosts

When do hosts trust an AI tip? Where do they look for listing content?

User interviewsCard sorting

Week 4

Synthesis

Which pains share a cause?

Cluster analysisAffinity mappingThe readout

Who I'd talk to

This is the plan I’d run with a team. A prep week re-scores the assumptions, opens data access and starts recruiting. Week one is data first, because exports need no recruiting. That means sales calls from Gong, support tickets, how hosts use field overrides and the Knowledge Hub today, and ten interviews with Hospitable’s own teams.

[Press] Hosts come second, sharpened by what the data shows. About twenty-five user interviews over two weeks, and a card sort with fifteen hosts. Each stage answers its own question, written on its card. Week four is synthesis and the readout.

The first questions go out on day one. Is the Booking.com text an API limit, or Hospitable’s choice? What are Airbnb’s write limits? How can Hospitable spot drift? I simulated the host side to show the method. I did not simulate what Integrations or Legal would say. Those stay as nineteen open questions, each with an owner. About 55 seconds.

ResearchMethod

What I'd research, and what I already have

I'd answer the question with 25 user interviews, a card sort and the data, and I already have public evidence.

1The question

Do hosts vary money and stay rules per channel, or only the words?

2Who I'd talk to

25

user interviews, including hosts who churned or never joined after a trial

15

hosts in a card sort

3What I'd check

618

sales calls in the brief

300

support tickets to tag, plus 100 untagged

4What I already have

115

reviews and posts

89

channel concepts

36

competitor products

5How an insight gets strongStrong means three sources of two kinds, one of them real

Interviews, card sort and ticketsSales callsPosts and reviews at trustpilot.com, channel docs and competitor help centres

ResearchPersonas

Hospitable is the source of truth, earned, not declared

Hosts keep their truth where it's easiest: Airbnb, a doc, a sheet. Hospitable is only trusted when every channel matches it.

Stock photo illustrating Priya, not a real host

“You're the messages.”

PriyaNew to a PMS

One channel, fears reviews most

Time spent on listings, monthly

None, “that's the point”

Stock photo illustrating Marta, not a real host

“On Airbnb”

MartaFamily host

Varies only money per channel

Time spent on listings, monthly

About 1 hour

Stock photo illustrating Dana, not a real host

“My doc”

DanaSEO optimiser

Tunes titles, openings, photo order

Time spent on listings, monthly

6 to 8 hours

Stock photo illustrating Callum, not a real host

“The sheet”

CallumPortfolio manager

Compliance, markups, rate plans

Time spent on listings, monthly

About 20 team hours

The four host typesInterviews with four hosts, simulated; photos are Unsplash stock, illustrative

ResearchInterview synthesis

Hosts can't see where a change lands

Who

1Scattered

Nobody's real version is Hospitable's.

The AI speaks from a store no channel sees.

2Unclear where it lands

Saving says nothing about where it went.

They want to see a change before it's written.

A change shown as a field is hard to picture.

3Each channel differs

Money and stay rules differ per channel, not words.

Copy per channel is one persona's job.

On Booking.com, the voice is the facts.

SoA canvas. Each channel's real guest page, with the change on it, before save.

Five simulated interviews, grounded in public reviews

This is the synthesis, and it matters most in this section. Five simulated hosts told me three things. First, listing management is hard because it’s scattered. The real version lives on Airbnb, in a document or in a sheet. Second, it’s unclear where an update lands. Saving says nothing about where it went. Third, each channel is different, so one strategy doesn’t fit.

[Press] Choose a host to see what matters to them. The cards that aren’t theirs fade, and hovering a chip previews it.

[Press] Callum and Marta both want to see a change before it’s written. Marta says she gets nervous picturing it on her real flat. That’s why I push for a canvas. It shows each channel’s real guest page with the change on it, before you save.

[Press] The four needs behind it are in Notion. See it before it’s written. Know where it went. Handle exceptions in bulk. And block invented facts. Everything here is simulated, so a real round of about 25 interviews has to test it first.

ResearchPain points

Four pains the research found, each with a product decision

Hosts keep the truth elsewhere, can’t see changes, and distrust AI that sounds surer.

1

The truth lives somewhere else

Product decisionOne Base holds each fact once, and every channel inherits it.

2

Per channel means money and rules, not words

Product decisionMoney and stay rules can differ per channel. Copy stays shared.

3

Changes land with no preview

Product decisionEach change shows its reach first, then a receipt confirms it.

4

The AI sounds surer than it is

Product decisionEach suggestion says why, where and how to undo it.

Sampled interviews and support dataHospitable’s public feedback board, where hosts ask for what they need

ResearchInsights

Six insights shaped Smart Listings, each traced to its evidence

Six insights, by signal

51

Explaining raised trust, not checking

In simulated sessions the AI wrongly said the flat had a balcony. Explanations made people accept more, not check more. So invented facts must be blocked.

The insight that changed the design most

Per channel means money and stay rules, not words

Only the optimiser wants different words per channel: a channel's voice is a choice

They want to see a change before it's written, and proof after

The Knowledge Hub repeats the listing, and misses what guests ask

Synthesis of interviews and data, checked against the real product and public evidence

I grouped the sampled interviews, data and card sort with what I could see for real. Six insights shaped Smart Listings: five Strong, one Moderate. Per channel means money and stay rules, not words. Only the optimiser wants different words per channel. So a channel’s own voice is a choice, not a default. That’s why a field uses the Base until the host customises it, and why Smart Listings is opt-in. Hosts want to see a change before it’s written, and proof after. And the Knowledge Hub repeats the listing in its own copy. The guest AI answers from that copy, while nothing learns from what guests actually ask.

[Press] This is the one that changed the design most. In the simulated sessions, a why next to an AI suggestion made hosts trust it more, but not check it more. Only two of five noticed when the AI wrongly said the flat had a balcony. So the AI may only say what the facts back. That has to be a hard rule, not a warning. And nothing here proves hosts want AI copy. About 55 seconds.

ResearchJobs to be done

One goal, four jobs: a listing that's right, and checkable

Journey map for changing a fact once, across six stages, with today's functions and gaps

Journey map for setting money and stay rules per channel, across six stages

Journey map for writing for each channel, across six stages

Journey map for confirming what went where, across six stages

Today's functions and gapsHelp CenterSmart Listings on sample data

ResearchCard sorting

I want to test where listing content belongs

Sort these cards into the groups that make sense to you, or make your own. A card sort tests how hosts think about their listing.

Cards to sort

Photos
Summary
Title, description
Getting insight and setup
Guest resources
House rules, Wi-Fi
Location
Address, neighbourhood
Amenities
Parking, safety
Booking policy
Cancellation, minimum stay
Pricing
Availability
Calendar

Shared across channels

In one channel

A group of your own

Name it, then drop the cards that fit it here.

Card sort, not run yetHospitable's property menu, my.hospitable.com

ResearchChannel constraints

Every channel sets its own limits

Booking.com writes its own description, so one editor can’t cover every channel. It is one example of a limit, and each channel has its own.

Airbnb

Hospitable writes it

An edit made on Airbnb forks the field

Writes its own

Booking.com

Booking.com writes it

From structured facts. Hosts can only request a fix

Vrbo

Hospitable writes it

Edits made on Vrbo are overwritten

Expedia

To be mapped

Its limits are part of the planned research

Direct

Hospitable owns it

No limits on title or description

Channel	Title	Description
Airbnb	≤ 50 characters	Sections, no published limit
Booking.com	3 to 255	Written by Booking.com
Vrbo	20 to 40, banned words	One field, 400 to 10,000 characters
Expedia	To be mapped	To be mapped
Direct	No limit	Editable per channel

Each channel’s limits, to be mapped in researchThe channels’ own documentation, Booking.com as one exampleThe limits

ResearchCompetitive analysis

Tools compare prices. Almost nobody compares the listing itself.

Pricing tools check your rate against similar listings. The listing itself goes unchecked.

1

Comps for price, today

Generic recreation

Nightly price12 similar listings

Your price€142

Similar median€128

You

€96Middle half€180

4 of 6market tools price by comps

2

Comps for the listing

Concept

Your listing12 similar listings

	You	Similar
Title length	31	46
Photos	14	27
First photo	Bedroom	Living room
Amenities named	4	9
Description sections	2	5

Photos

Similar listings lead with the living room photo

Why

9 of 12 open on the living room.

2 of 36products point comps at the listing

SoAI comps for the listing, per channel, as Copilot tips.

Comps for price: PriceLabs and Beyond’s product pages, pricelabs.co and beyondpricing.comBoth panels are my recreations, with demo numbers

Every pricing tool I reviewed works the same way. PriceLabs, Beyond, AirDNA and Wheelhouse build a set of similar listings on the same channel. Then they put your nightly price against it: your rate, the median, and the middle half of the market. Four of the six market tools do this.

[Press] Nobody does that for the listing itself. The same twelve similar listings could show a host that their title is shorter, they have half the photos, and they name fewer amenities. Only AirDNA’s amenities list and Rankbreeze’s keywords come close, and only in part.

[Press] That’s the opening for Hospitable’s AI Copilot. Read the similar listings and turn the gap into a tip, like leading with the living room photo. Per channel, because similar listings on Airbnb aren’t the same ones as on Booking.com. The numbers are only illustrative. About 45 seconds.

ResearchProblem statement

The problem: Hospitable gives hosts no listing they can trust

The problem

Hospitable keeps a home’s facts in two copies, varies the wrong things per channel, never says where a change went, and offers AI that persuades instead of being safe.

How might we…

1
Two copies
keep one record of the home that every channel and the guest AI read?
2
Wrong per channel
give money and stay rules one per-channel pattern, and make voice a choice?
3
Unseen changes
show each channel’s listing, and say where every change went?
4
Persuasive AI
let the AI write only what the record can back?

The product auditMy synthesis of the sampled researchFour root causes

ResearchObject model

Every pain behind Smart Listings is a missing object

An object model is a product's nouns: what it's made of, and how they connect.

Property

The home. Every fact, once.

Listing

The home on one channel.

Sync direction
not yet established

Source

Where each fact came from.

Projection

One fact, worded for each channel.

Copy

What guests read: shared, or per channel.

Receipt

Proof a change reached the channel.

A fact comes from a Source, lives once on the Property, is projected to each channel, and becomes the Copy on a Listing. Every save returns a Receipt.

From the briefNew: 4 of 22, missing today

Property, listing and the sync directionThe whole model

ResearchChannel constraints

One bathtub, four names. The host states it once.

A host shouldn't need four dictionaries to say they have a bath. Hospitable does the translating.

The host says it once

Bathtub

Value: Yes
Scope: Bathroom 1

Hospitable translates it for every channel, as it already does for amenities

AirbnbAmenity "Bathtub"
Booking.comRoom facility BATH, per bathroom
VrboTUB, full bathroom only
ExpediaBATHTUB_ONLY, a shower type

Airbnb, Booking.com, Vrbo and Expedia documentation · Hospitable's help centreWhat it forces

This is the principle behind everything that follows, in its smallest example. It shows why one editor per channel can’t work. Here is one fact, the way a host would say it: there’s a bathtub in bathroom 1. Now watch where it has to go. Airbnb files it as a simple amenity. Booking.com files it as a room facility, per bathroom. Vrbo only counts it in a full bathroom. Expedia files it as a kind of shower for the whole unit, so which bathroom it’s in is lost.

[Press] A host should say yes, there’s a bathtub, once, and never think about it again. Hospitable already does this translation for amenities. The host sets them once, and Hospitable pushes them to each channel. The shared layer extends that to every fact, one of a hundred standard fields. Today the product has only two states per field: one value, or Edit per channel. About 30 seconds.

ResearchNavigation options

Five navigation models. I chose Channel-first.

I drew five ways Smart Listings could fit the menu, and compared them on the same tasks.

1Mode switch

A switch swaps the whole menu

Focused workspace

Other sections vanish

The review’s pick

2One listing hub

One hub, a channel selector inside

No second copy

Biggest migration

3Sub-nav section

Today’s menu, plus one group

Nothing moves

Two copies drift

My pick

4Channel-first

Shared, then each channel’s pages

Fewest clicks

Grows per channel

5Assistant drawer

A drawer on every page

No menu change

No room, no home

SoShared holds each fact once. Each channel’s pages are where it can differ.

Five menus drawn in Paper and reviewed on the frames, not tested with hostsThe menu

ResearchSitemap

Shared traits once. Channel actions where they differ.

Laying out the sitemap, I saw hosts need to know what each channel repeats from the Base.

What each channel takes from the Base

Uses the Base
Custom
Channel only
Set by the channel

Airbnb2473

Booking.com2542

Direct31

On every field

Summary

Your quiet base for working remotely…

Custom for AirbnbUse the Base

Summary

A sunny two-room flat…

Uses the BaseCustomize for Direct

Counts: the listing fields of my prototype's sample flatThe trade-off

ConceptDesign principles

Seven principles, each one able to lose an argument

I let the research rewrite my draft. Three changed, one was added, and number three became the spine.

Design principles 7

1Nothing written unseenNothing changes until AcceptRewritten
2Facts onceOne value per detailNew
3Money and rules vary per channel. Voice is a choiceFees and minimum stays vary per channel, words rarelyRewritten
4The AI says only what the home can backEvery tip names its sourcesRewritten
5Autonomy follows the stakesMoney always waits for Accept
6One product for four hostsOpt-in per property
7Say where, and whether it arrivedReceipts show Sent, then Live

My draft		After research	Why
Nothing is lost	→	Nothing written unseen	A restore point answers the wrong moment
None	→	Facts once New	Two stores of facts disagree
Facts shared, voice local	→	Money and rules vary. Voice is a choice	Money and rules vary per channel, words rarely
Show the why first	→	The AI says only what the home can back	A reason raises acceptance, not scrutiny

From the simulated researchWhere each came from

ConceptDivergent ideationRound one

Seven mental models, and the two that made a pair

Each had its own primary object and first action, all drawn on the same flat. None won every screen, so I kept two that cover each other.

Channel Canvas

“Check my shop window”

What will guests see?
Golden Record

“A master spreadsheet”
Copilot

“Tell someone what you want”
Decision Inbox

“Only show me what needs me”

What needs me?
Merge & Versions

“Drafts and versions”
Fact Blocks

“A house manual I write once”
Home Tour

“Walk me through your flat”

Scores are my design judgement, made before any concept testFrames and scores

Round one asked how a host could think about a listing at all. I wanted directions that were truly different. So I set a hard bar: no two could share a mental model, a main object or a first action. Here they are, each drawn as the thing you’d work on. The channel’s own page. A master spreadsheet. A conversation. An inbox of decisions. A version history. A house manual. A floor plan. Every frame used the same flat in Hospitable’s design system, so they could be compared.

[Press] No model won every screen. The Channel Canvas scored best where a host shapes each channel’s page. The Decision Inbox scored best where a host runs many homes. So I kept them as a pair. One answers "what will guests see?" The other answers "what needs me?" The frames and my scores are one click away. The scores are my judgement, made before any concept test, so they’re not host evidence. About 45 seconds.

ConceptPattern referencesRound two

Twelve AI patterns, each borrowed from the best

I ran the same Smart Listing flow through twelve patterns people already trust. Two became the core of the build, and four lent it a part.

In the buildA part of it

Listing score

Google Ads · Yoast
Live canvas

Squarespace · Notion
Copilot on every screen

Shopify Sidekick · Notion
Channel composer

Buffer · Hootsuite
What-if simulator

Credit Karma · PriceLabs
Strategy picker

Grammarly · Spotify
Goals and guardrails

Performance Max · Advantage+
Suggesting mode

Notion · Linear
Readiness guide

Squarespace · Vanta
You vs similar homes

Hootsuite · Tripadvisor
Rate calendar first

Turo · KAYAK
Guided pass

Zoho CRM · Navattic

Pattern references from the vendors’ own pages and saved screensSame flow

Round one gave me a structure. Round two was about the listing itself. How should AI help a host make each channel’s listing better? I built twelve directions in Paper. Each one uses a pattern people already trust, from Google Ads’ optimisation score to Notion’s suggesting mode. I saved the references beside each one. Borrowing what hosts already know means less to learn. Every direction ran the same flow, from the overview through pricing, availability and policies to receipts. And the same hard rules applied. Money and stay rules never go on autopilot. An unknown fact becomes a question, never a guess.

[Press] Two became the core of the build. The live canvas is now the Canvas view. The copilot became the docked tips panel and Rewrite with AI. Four lent a part: who the strategy is for, its goal and guardrails, the tracked change with its why, and similar homes as evidence. About 50 seconds.

ConceptFinal concept

Two rounds of directions, one model: Smart Listings

Round one set the shape. Round two gave the model. A strategy per channel turns into tips with their why, seen on the channel’s real page.

Round one

Round two

In Smart Listings

Strategy picker
Goals and guardrails

A strategy per channel

Who it’s for, a goal, guardrails

Decision Inbox

Copilot
Suggesting mode
You vs similar homes

Tips with their why

One at a time, with sources

Channel Canvas

Live canvas

The Canvas view

Each channel’s real page

Copilot

Rewrite with AI

Any field without a tip, even with Smart Listings off

How it converged

Here’s how two rounds became one model. Round one gave me the shape. The inbox’s "what needs you" became tips. The canvas became a window on each channel’s real page. Round two gave the parts: the live canvas, the copilot, two strategy directions, the tracked change, and similar homes as evidence.

[Press] Put together, that’s Smart Listings. A strategy per channel: who it’s for, a goal and guardrails. Tips with their why, one at a time, in a docked panel. The Canvas view of each channel’s real page. And Rewrite with AI on any field without a tip. It needs no strategy, so it works even with Smart Listings off. I reviewed the live canvas frame by frame, with 43 comments, and that review set the model. When the strategy grew to four steps, I cut it to one question per screen, on Hospitable’s own wizard. Then I built it for real. From here on, everything I show is running code. About 55 seconds.

DesignSolution overview

One listing for every channel, and tips that say why

Facts live once, channels keep only what differs, and you decide what the AI does alone.

Property name

Name a landmark guests search for

Creates Airbnb’s own value. The Base stays as it is.

~~Beautiful Artist Apt. in Neukölln~~Artist's Flat by Tempelhofer Feld · Neukölln

Why

Remote workers search by area first: 17 of 24 similar Airbnb homes nearby name the neighbourhood and a landmark.
Tempelhofer Feld is 5 minutes on foot.

Shared: every fact once

Channels: one page set each

Every field shows its source

A tip, with its why

Tips and reasons, demo dataThe map, and what I cut

DesignListing base

Change a fact once, and see every channel it reaches

Every channel and the Knowledge Hub read the same listing detail. So the Listing base holds only what channels share, and each field shows its reach.

Listing base · Arrival

Check-in time

Kept once, here.

Shared with every channel that reads it

Read byAirbnbBooking.comDirectKnowledge Hub

Labelled Source of truth
Each field shows the channels it reaches
Channel-only fields move to their channel’s page

Property name reaches 2 channels, the summary reaches 1 channel

Source of truth

Its reach, under every field: never a surprise

Product audit: the check-in time from a trial listingThe prototype, on demo data

DesignKnowledge HubConcept

The Knowledge Hub learns from guests’ questions

A concept, not built in full: guests’ questions reveal gaps, and each gap becomes a tip for the Base or one channel.

Inbox

Is there free parking?

Knowledge Hub

QuestionsInsightsFactsSources

Is there free parking?Asked 9 timesNot answered
What time is check-in?Asked 14 timesAnswered

Amenities

Add free parking

Why

Asked by 9 guests, not answered

Listing base

Free parking on premise

Guest AI

1A guest asks in the Inbox

2Counted, answered or not

3A gap becomes a tip

4The host’s answer becomes a fact

5Every channel and the AI use it

Product audit: guest questions stay in the Inbox todayA concept: the questions and counts are demo dataThe flow

Today the Hub has no job of its own, so I gave it one: guests’ questions. It no longer keeps a copy of the listing. It reads the listing details from the Listing base and adds the situations the host writes. It has four sections: Questions, Insights, Facts and Sources. It starts in the Inbox. A guest asks if there’s free parking, and the question lands in the Hub. It’s grouped, counted, and marked answered or not. This one was asked nine times and never answered. The guest AI only answers from what backs it. When nothing backs an answer, it says so and asks the host.

[Press] That gap becomes a tip, with the same tip format as everywhere else. Its why is the questions behind it. It goes to one of two places. A shared fact for the Listing base, like free parking. Or a better answer in one channel’s listing, like Airbnb’s description.

[Press] When the host accepts, the answer is saved where it belongs. A listing detail goes in the Listing base, so every channel and the guest AI use it. A situation goes in the Hub. Either way, the next guest gets an accurate answer. To be clear, the Hub is a concept, not built in full. It shows how the Hub can feed the tips. The counts are demo data. The full flow is one click away. About 45 seconds.

DesignChannels

Optimise each channel on its own, starting from the Base

The Base is where a channel starts, not where it stops. On every field, the host sets how much each channel takes from it.

Follows the Base

The same on every channel.

Airbnb · Listing

Other details to note

Uses the Base

Customized for Airbnb

Its own words, for Airbnb search.

Airbnb · Listing

Summary

Custom for Airbnb

Only on Airbnb

No other channel asks for it.

Airbnb · Listing

Check-in method

Airbnb only

The prototype’s demo flat, as builtSet by the channel

DesignStrategy

Turn it on, then aim each channel in four questions

Opt-in per property, since not every host wants AI. It reuses Hospitable’s Add-property wizard. Each suggestion says why.

Goal

Who it’s for

Guardrails

Autopilot, off by default

Airbnb · Strategy 1 of 3

What should this strategy aim for?

Fill empty nightsSuggested8 of your 13 empty nights on Airbnb are Monday to Thursday. Earn more per bookingYour Airbnb weekends book 3 weeks out. Similar homes charge €12 more.

Set your guardrails

Smart Listings never goes outside them.

Keep my nightly price between

€

to

€

Prices and fees, Minimum stay and rules, Cancellation policy · Always asks you. This can’t be turned off.

Pattern: Hospitable’s Add-property wizardSuggestion evidence, demo dataThe three strategies

Smart Listings is opt-in, switched on per property, because not every host wants AI on their listing. With it off, Shared and Channels look exactly the same, so the structure serves every host. Turning it off later keeps everything. Turning it on opens a wizard hosts already know: Hospitable’s Add-property flow. So I reused it rather than invent a new pattern. First, what to aim for. Each option shows its evidence, like this one: 8 of your 13 empty nights on Airbnb are Monday to Thursday.

[Press] Then the guardrails, which are the contract. Smart Listings never goes outside the price range. Prices, stay rules and cancellation always ask. That lock can’t be turned off.

[Press] Last, whether Autopilot may act. It starts off. There’s one strategy per channel, or one shared by several. So Airbnb and Booking.com can go after different guests. About 55 seconds.

DesignTips & AI rewrite

Accept a tip knowing exactly what changes, and why

A suggestion without a reason is a guess. Each tip sits beside its field, with its sources.

The edit, shown once
Every reason, sourced
Undo, one click away

The Airbnb listing with the Wi-Fi tip open, as built in the prototype

The space

Say how fast the Wi-Fi is

…by the window with an ergonomic ~~chair and~~ chair, a 27" ~~monitor,~~ monitor and 250 Mbps fibre Wi-Fi (tested on 12 Sep), plus the paintings, a sofa and…

Why

From your Knowledge Hub · Wifi/Internet: 250 Mbps fibre, tested on 12 Sep.
5 of your last 12 Airbnb inquiries asked about the Wi-Fi speed.
8 of 24 similar homes state the speed in The space.

Tip content and reasons, demo dataOne AI per field, and prices tooThe flow

Today the only listing AI in Hospitable is a separate page of suggestions, away from the fields. A suggestion without a reason is a guess. So I put every tip beside its field, in one docked panel, one at a time. This card is the real component. The tip is to say how fast the Wi-Fi is. It’s the fourth of eleven. It shows the edit once, with only the new words coloured.

[Press] Every reason names its source: a fact from the Knowledge Hub, the host’s own inbox, or similar homes nearby. Hover the icon and it says which.

[Press] Accept is Hospitable’s usual raspberry button. Let me accept it. And Undo is one click. One rule kept this calm: a field shows one AI marker at most, either its tip or Rewrite with AI. Rewrite with AI doesn’t need Smart Listings. Every host has it. About 50 seconds.

DesignCanvas preview

See the listing as guests will, then tap to change it

Hospitable publishes with no preview today. I rebuilt each channel’s real guest page. A tap opens the field behind it, its tip or Rewrite with AI.

Purple: a tip waits here
Orange: editable here
No colour: owned elsewhere
Show only editable sections: hide the rest
Your edits or Live now: compare, and go back to the original

Canvas: the Airbnb page as guests see it, with the title tip open

Canvas: the Booking.com page as guests see it

The title on Airbnb's page, ringed in purple: a tip waits here

Classic or Canvas, one switch

Changes publish without a preview

Canvas goes beyond a preview. One switch shows only the sections you can edit. Another switches between your edits and what’s live on the channel now. So you can see the difference, and go back to the original with one click. Today no screen in Hospitable shows one listing as it looks on one channel, and changes publish without a preview. So Canvas shows the channel’s real guest page, and colour says what you can do. Purple: a tip waits here. Orange: you can edit it here. No colour: it’s owned somewhere else. On Airbnb the title has a purple ring, because a tip waits on it.

[Press] Booking.com gets its own page, one switch away from Classic. Canvas works for every host, with Smart Listings on or off. My first version let you edit right on the page. In review, that became a second editor for the same field, so I took it out. Now a tap opens the field itself, and a price opens Pricing’s own fields. One field, one editor. About 50 seconds.

DesignPerformance · concept

Each strategy, measured against the weeks before it

A concept: since the strategy started, against the same weeks before.

Airbnb occupancy

+20points

60% since 14 Aug, 40% before

Airbnb revenue

+39.2%

€1,507 since 14 Aug

OccupancyShare of nights booked through Airbnb.

7 weeks before 40%Since 14 Aug 60%+20 points

Strategy started 14 Aug

25 Jun14 Aug25 Sep

Next ideaStrategy versions on Airbnb

Version 1Since 14 Aug · +20 points

vs

Version 2From the next change

Performance, demo dataCards and bars follow Hospitable MetricsStrategy versions

DesignAutopilot

Autopilot for words and photos. Money always asks you

With Autopilot on, text and photo-order tips apply inside the host’s guardrails, since they are cheap to undo. Prices, stay rules and cancellation always wait for a yes.

May apply on its own

Text tips
Photo order, never adding or removing a photo

Always asks you first

Prices and fees
Minimum stay and rules
Cancellation policy

Always asks you first: prices and fees, minimum stay and rules, cancellation policy. Autopilot on

Pricing: Autopilot is on, prices always ask you first

Locked: it can’t be turned off

The editor says Autopilot is on

The Autopilot question

DesignMobile & tablet

Fix a listing between turnovers, from your phone

To make that possible, I built an app mode that runs every screen as Hospitable’s mobile app: sheets, swipe-back, tips in a bottom sheet.

Every side panel becomes a bottom sheet
44 px targets, 16 px inputs
Checked at 375, 768 and 1440
Installable: Add to Home Screen
Voice input: a stand-in, no speech model yet

App mode: the Wi-Fi tip in a bottom sheet

Built in part, under the case’s time limit. Many improvements are still needed

A caveat first. I built this only in part, because of the time I had. It would need many more improvements before it shipped. I designed for a host fixing a listing between turnovers, on a phone. That’s my design intent, not a finding. It means mobile isn’t a shrunken desktop. Here is the Wi-Fi tip again. On a phone, every side panel, tips included, becomes a bottom sheet you can drag away, with 44 pixel targets and 16 pixel inputs.

[Press] I built it so the same code runs as an installable app: Overview, Performance, and every screen I settled on. Pages slide in and out, and you swipe back from the edge.

[Press] Both the website and the app are live, so let me open the app and show it. Voice input is a stand-in. It shows the flow, but there’s no speech model behind it, and I say so. About 45 seconds.

DesignPresets

Start from defaults that fit you. Change any of them

The four hosts in my research want different amounts of help, not different products. So I designed presets, not personas: defaults change, capability never does.

Host	Preset
P1 Family host	One home · Standard guidance, Smart Listings off
P2 SEO optimiser	Few homes · Expert detail, Smart Listings on
P3 Portfolio manager	Portfolio · Standard guidance, Smart Listings on
P4 New to a PMS	One home · Guided, Smart Listings off

“We don’t label you”

Personas from the simulated research

DesignDesign system

Three patterns I challenged

They change how every part of Hospitable is organised, not only listings.

Drill-down navigationOnly the main menu had it. Now every section does.

Counts that moveThe total on the parent when closed, each page’s own when open.

Property name

Guests see it in search.

Beautiful Artist Apt. in Neukölln

Stacked fieldsLabel above the input, so Tips or Canvas fit beside it.

Also: Save changes only when something changed, and one outlined ×.

Built on Hospitable’s own components, as in the prototype

DesignDesign system

Patterns I introduced, and where they reach

Before they spread, I’d weigh them with each team, then test.

Wi-Fi

Say how fast the Wi-Fi is

DismissAccept

Tip card

Tips · 3 2

Tips drawer, markers

Description

Rewrite with AI

Uses the BaseCustomize for Airbnb

Channel badges, reach caption

ClassicCanvas

Canvas preview

Changes savedUndo

Save toast, Undo

Essentials5Arrival2Location2

Section tabs with tip counts

Reaches

Inbox
Calendar
Pricing
Operations
Guest experience
Metrics
Mobile app

Evaluate, then test

Built on Hospitable’s own tokens and componentsThe areas they reach are my reading, to checkThe impact

TestingTest plan

Real hosts test it next, against today’s product

Nothing is tested with real hosts yet. Six steps follow, each with a bar set before it runs.

Benchmark todayToday’s product, 40 hosts
Sets today = 100
Tree testThree menus, 30 people each
≥ 70% on the channel tasks
Usability, 2 roundsThink-aloud, 8 hosts a round
Time to adjust ≤ 75
A/B testToday against channel-first
Faster. No task loses over 10 points
Pilot, six weeks30 accounts, a private Slack
≥ 80% satisfied
Staged rolloutWith 10% held out
Beats the holdout

Same tasks at every step, so results compare. Accessibility gates each one.

My test plan: every bar is written before its step runsThe probes and the bars

TestingGoals & KPIs

Every goal tests one of my riskiest bets

The goals I set at the start, each with its KPI and test.

North star

Time to adjust ≤ 60today = 100

Benchmark, A/B

Trust check

Satisfied ≥ 80%

KPI

Bet

Test

+7%

More bookings and requests

Bookings per listing

H-10Channel copy sells

Rollout vs holdout

25%

Hosts adopt it

Properties with a strategy

H-04A voice per channel

Pilot, per segment

50%

Hosts keep refining

Adopters acting monthly

H-09Trust in AI tips

Pilot, six weeks

40%

Tips earn their place

Acceptance rate

H-09Trust in AI tips

Usability rounds

−10%

Fewer losses, less support

Tickets per 100 properties

H-15What Autopilot may touch

Rollout vs holdout

Targets at 90 days, set before any data. The pilot sets the real onesThe bets, from my assumptions mapThe KPI treeHow it could fail

Here is how I’d know it worked, tied back to where I started. One north star: time to adjust. That’s how long a host takes to make a real change correctly. Today’s product sets 100 in the benchmark. The target is 60 or less. Satisfaction at 80 percent is the trust check beside it.

Under it are the five goals from the start, each with its KPI. Bookings per listing up seven percent. A quarter of properties with a strategy. Half of adopters acting every month. Forty percent of tips accepted. And ten percent fewer tickets.

[Press] And each goal tests one of my riskiest bets. Trust in AI tips sits under acceptance and refining. Usability testing, then the pilot, check it. Who wants a voice per channel sits under adoption, read per host type in the pilot. What Autopilot may touch sits under losses, read against the holdout. If a bet fails, the goal shows it first. About 45 seconds.

Next stepsStory and product

“Your listings, handled. And you still have the last word.”

1What I’d tell hosts

One Base

On Airbnb

DescriptionUses the Base

SummaryCustom for Airbnb

Check-in methodAirbnb only

Your facts, once

AI Copilot

Property name

Name a landmark guests search for

Why

Tempelhofer Feld is 5 minutes on foot.

Tuned for each channel

Receipts

Property name updatedApplied

Weekly discountAsks you

Money always asks you

2Where I’d take the product

The property around one Base
Copilot wherever the listing goes
Smart Listings across Hospitable

Mapped next

AI Copilot: hospitable.comThe product direction is a plan

Next stepsService map

Smart Listings is a system, not a screen

Facts kept once and one kind of tip can reach every part of Hospitable.

BuiltNextLaterImplied, not built

Shared

Smart Listingsstrategy · tips · Copilot

Listing base

Knowledge Hub

Airbnb · Booking.com

Metrics

Direct sites

Inbox · guest AI

Similar homes, per channel

Tips in the Calendar

Performance over time

Import

A deeper Knowledge Hub

Tips in the Inbox

Owner statements

Dynamic Pricing

Google

Guidebook

Operations

Devices

API · MCP

Built

tips · the Base results the Base answers price, positioning tips periods facts, once questions tips receipts goals answers, insights

Built: in the prototype, on sample data. Next and later: my plan, nothing measured.Where it falls shortPerformanceRoadmapWhat it sets off

Smart Listings only pays off as a system, not a screen. In the middle is what I built: Shared, which is the Listing base and the Knowledge Hub that reads it, with Smart Listings on top. The inner ring is in the prototype. Tips and the Base reach Airbnb, Booking.com and Direct. The guest assistant answers from the same facts. And results come back as Performance.

The next ring is where the build falls short, and I want to name it. Four lines of work are implied but not built. One: similar homes for each channel. Two: a deeper Knowledge Hub whose job is guests’ questions. It collects them from the Inbox, answers only from facts, shows what guests ask, turns gaps into tips, and knows which messages and tasks use each fact. Three: Performance a host can act on over time. Four: tips beyond the listing, in the Calendar and the Inbox first, all with one tip format. Import joins here, on the same Base.

The outer ring is later. Pricing reads the strategy’s goals. Owner statements read the receipts. And the Base’s facts and the Hub’s answers and insights feed the Guidebook, Operations, Devices, Google and the API. How I’d run each line of work is one click away. About 70 seconds.

Close

On Monday, I’d swap simulation for real answers

Four things, one week. Every team gets its questions in writing. Friday’s readout decides what changes.

1Benchmark today’s product
2Ten stakeholder interviews
360 sales calls, two coders
4Ticket and edit data

MonTueWedThuFri Me Product Integrations AI/ML Data CS Legal Hosts

2Kick-off. Questions out

2Owners and roadmap

4Data access

1Benchmark set up

2What each channel accepts

2What feeds the Hub

2Overrides, signals

4400 tickets

1Recruiting with Marketing

360 calls coded

4Usage exports

1Benchmark: 40 unmoderated, 5 moderated

4Tickets, edits read

2What needs a person

2AI copy, channel terms

ReadoutOne page to every team: what I learned, what changes

1Round 1 booked: 8 hosts

My plan, not run. Sizes from the research plan and the test plan

Everything I’ve shown is designed and built, and the evidence about hosts is sampled. So on Monday I’d start making it real, with four things: a benchmark of today’s product, ten interviews with Hospitable’s own teams, the sales calls tagged properly, and the ticket and edit data. Monday is the kick-off with the Head of Product and the listing team. Every owner gets their open questions in writing. [Press] Tuesday, Integrations tells me what each channel accepts. AI and ML tell me what feeds the Knowledge Hub. Data tells me how often hosts override fields. [Press] Wednesday, sixty calls get tagged, checked by a second person, and the benchmark runs on today’s product. [Press] Thursday, CS and Legal. [Press] Friday, a one-page readout to every team, and round one of usability testing is booked with hosts.

Thanks.

hospitable

Managing listingsacross channels.

Smart Listings

If every channel gets its own strategy, and recommendations backed by data, listing management stops being a chore and becomes Hospitable's competitive advantage.

1Scattered

2Unclear where it lands

3Each channel differs

The truth lives somewhere else

Per channel means money and rules, not words

Changes land with no preview

The AI sounds surer than it is

Similar listings lead with the living room photo

Why

Design principles 7

Name a landmark guests search for

Why

Add free parking

Why

What should this strategy aim for?

Set your guardrails

Say how fast the Wi-Fi is

Why

Say how fast the Wi-Fi is

More bookings and requests

Hosts adopt it

Hosts keep refining

Tips earn their place

Fewer losses, less support

Your facts, once

Name a landmark guests search for

Why

Tuned for each channel

Money always asks you

Similar homes, per channel

A deeper Knowledge Hub

Shorten the minimum stay for this gap

Why

Add the answer guests keep asking for

Why

Test and harden it

Import, and what it implies

Optimise, then go wide

Blast radius

A channel changes its terms

Comparing with similar homes

New support questions

Thanks.