Retrosuperfuture × defil.ai
01 / 20
Retrosuperfuture × defil.ai
Brainstorming session · indicative numbers

See yourself in RSF — and never leave a customer unanswered.

Two builds. One reinvents how a shopper tries on a frame — from the camera to a moving picture of themselves in their own world. The other gets the routine done on your one customer-service desk.

2 use cases Shopify-native · inside your platform camera → look → motion brainstorm · 21 Sep 2026
Prepared for Filippo Baroli — CEO, Retrosuperfuture
Indicative numbers — two unknowns priced as placeholders.
The core advantage

A blank canvas — the right infrastructure from day one.

Where RSF starts
no legacy stack one desk a shop window to reinvent
An AI-native RSF
built on the right rails
AI Immersive Try-On → sales ↑
E2E AI Customer Service → costs ↓
= more sales · lower costs
We built inside big companies — and know how slow they can be
Luxottica Logitech FedEx Credit Suisse HKSAR Government Here we can be FAST
No committees, no legacy: decide in a call, ship in weeks.
The goal of this engagement

Make the shop window disruptive again — and take the routine off the one desk that answers everyone.

Use Case 01 · sales + Use Case 02 · operations = more sales, lower costs

One goal, two builds: sell more from the shop window, spend less answering the routine.

What we heard from Filippo

A new shop window. One desk answers everyone.

Use Case 01 · AI Immersive Try-On
Not an overlay — a render of you.
Fit, colour, scene, motion — from the packshots.
Use Case 02 · E2E AI Customer Service
From three people to one desk.
Opticians, distributors, shoppers still write there, every day.
« Sarebbe fighissimo avere un virtual try on un po' diverso… come quando vai a comprare una casa e ti fanno il render. »
“A try-on that’s different — like a house render.”
Filippo · our call · September 2026
« Prima di assumere un altro customer service, si può fare il piano di implementazione di un assistant AI. »
“Plan an AI assistant before hiring again.”
Filippo · the same call
This is my understanding — correct me where I’m wrong.
At a glance · what each build delivers

Two use cases. Two problems. One platform.

Use Case 01

AI Immersive Try-On

Camera, fit check, then yourself — every colourway, your outfit, your place, moving. Save, share, buy.
< 60 s
camera → moving look
3,900
frames, from your packshots
Removes “find a stockist” on 70% of variants
Use Case 02

E2E AI Customer Service

One assistant, three audiences, in your platform: routine resolved; your colleague approves only refunds, claims, documents.
70%
resolved without a person
3
audiences, one desk
Removes the wall of email before one person

Target outcomes · instrumented from day one · real numbers, not a guess

Use Case 01 · AI Immersive Try-On
01

From a webcam overlay to a picture of you.

Today
the usual overlay packshot on white find a stockist
with the try-on
Seven steps
camera fit colour background outfit place motion
Overlays show fit, not feeling; 70% of your variants end in “find a stockist”.
The shop window today · and where it goes

The frame sits on white — it belongs on you.

Today — the frame on white
Panamá Classic Havana — packshot
fit is a guess
colour is a gamble
70% → “find a stockist”
The shopper sees the object, not themselves.
With the try-on
Live camera — the shopper wearing the Panamá Classic Havana
camera
Same face, Panamá Black Ivory
colour
Same face and frame, Navigli, Milan
background
Same face and frame, a new outfit
outfit
LIVE · 5 s
motion
The shopper sees themselves, in their world — and a clip to share.
How it works · step through it

Seven steps, under a minute.

01
Open the camera
02
Check the fit
‹swipe›
03
Pick the colour
04
Change the background
05
Change the outfit
06
Choose the place
07
Bring it to life — save, share, buy
Step 1 / 7
Open the camera
One tap on the product page. The front camera opens in the browser — no app, no account, nothing stored until the shopper says so.
See it working · the tour runs itself, or click a step

The camera opens. Then you appear.

Live camera view — the shopper wearing the Panamá Classic Havana Same face, Panamá Black Ivory Same face and frame, Navigli, Milan Same face and frame, a new outfit Piazza del Duomo, golden hour
01 · Camera
AI-generated · your consent
Under the hood · camera
Front camera, in the browser. Nothing stored until the shopper says so.
Illustrative — a synthetic model from the Panamá packshots; in production, the shopper’s face.
At scale · every shopper, every step, at the same time

Not one shopper. All of them.

12looks in progress
1,284looks today
318shared
96add-to-cart
illustrative
in progress
done
log
Idea · not a promise
An idea to A/B test
At the last step: 15% off if you buy within 15 minutes — 1% less every minute.
15:0015% off
We’d test it together — offer, timing, who sees it.
Illustrative — four synthetic shoppers; in production, each card is a real person on the product page.
Why this works

Disrupt the eyewear shopping experience.

3,900
From your packshots
No 3D models — the whole catalogue, day one.
Identity preserved
Face and pose stay; frame, scene, outfit change.
RSF-curated scenes
Your creative team sets the palette.
A share engine
Every look: a page, a lead.
Disclosed & consented
AI-labelled, consent first, deleted on request.
Measured
starts→looks→shares→add-to-cart
BenchmarkA UK independent optician shipped a generative try-on and reports +20% frame sales. Peers (Lexxola, Akila, Chimi): no try-on.
Use Case 02 · E2E AI Customer Service
02

One desk, three customers — done, not drafted.

The problem
Three people became one. Shoppers, opticians and distributors still write to that desk — and every ticket is still opened by hand.
The idea
One assistant, three profiles, inside your platform. The routine resolved end-to-end; your colleague approves refunds, claims and documents.
Who writes, and what they ask

Same assistant. Three profiles. Three different answers.

Shoppers
online, in store
  • order tracking
  • customs duties
  • returns & exchanges
  • warranty claim, photos
  • spare parts
Routine resolved; refunds and claims need a person.
Opticians
1,335 doors, 22 countries
  • reorder ship date
  • backorders & ETAs
  • claim within 8 days
  • price list, images
  • display material
Backorders announced before they ask.
Distributors
by country, all brands
  • shipment status
  • customs: invoice, HS codes
  • CE documents
  • batch claims
  • stock & reorders
Document pack drafted; one click.
Knows who’s writing · replies in their language: IT · EN · FR · DE · ES · JA · KO · ZH
The overhead today · and where it goes

Three audiences, one inbox, one person.

Shoppers
Opticians
Distributors
…one inbox, one desk
Today — every ticket by hand
Three audiences
every language · every day · by hand
Your colleague
one desk
Even with drafts, every ticket is opened, checked, translated and sent by hand.
With the assistant
Three audiences
routine
Assistantresolves & follows through
decisions only
Your colleague
decides
The routine closes itself. Your colleague gets a decision — file already assembled.
How it works · step through it

From request to done — five steps.

01
A request arrives
02
Recognised — who, and what
03
Reads the right sources
04
Resolves or prepares
05
Your colleague decides
Step 1 / 5
A request arrives
Email, the site chat or the ticket queue — WhatsApp optional. Nothing changes for the customer: they write where they always wrote, in any language, at any hour.
See it working · click a request to switch

The customer asks. The assistant closes it.

support@retrosuperfuture.com · assistantreplies in the customer’s language · IT EN FR DE ES 日本語 한국어 中文
Request:
Illustrative examples — answers only from your orders, policies and templates; hands over when it should.
See it working · a live simulation

A queue of decisions, not a wall of email.

customer service · livesimulation · illustrative
Customers0 open
The assistant
98resolved by the assistant
28with your colleague
Your colleagueone click
Logeverything, timestamped
resolved by the assistantwith your colleagueopen
Illustrative — 70% resolved without a person; every case logged.
Why this works

Grounded in your records. A person decides.

A person keeps the last word
File ready, reply drafted.
refundsclaimsdocumentsone click
It read your rules
Warranty 24 months · returns 20 days · EUR 15 pro-forma.
Real records, read-only
Orders, tracking, shipments, stock.
Knows who’s asking
Shopper, optician, distributor.
Any language, any hour
IT · EN · FR · DE · ES · +3
Measured from day one
resolved without a personfirst replybacklog
BenchmarkA European eyewear e-tailer (B2C and wholesale) went from 25% to 75–79% resolved once its assistant read orders; industry median without that wiring: 45%.
Trust & data

Your data stays yours.

Shoppers & customers
faces · orders · money
consent first
The assistant
resolves the routine
read-only · one-click approval
Your platform
orders · shipments · stock
EU-hosted GDPR by design No training on your data Encrypted & scoped NDA first Deleted on request
One build touches faces, the other money — both are built to your rules.
The build plan · both use cases

Day 77: both live — the shop window disruptive again, the routine off the desk.

Workstream
W0W1W2W3W4W5W6W7W8W9W10W11
W0 · Sign NDA & SA — not counted
sign
UC1Frame prep & fit engine
frames → fit
UC1Scene studio
colour · background · outfit · place
UC1Motion & the look page
clip · look page · share
UC1UAT → live
sign-off → live
UC2Map & measure
read → numbers
UC2Shoppers, then trade
shoppers → approval screen → trade desk
UC2Shadow mode
proposes · approves
UC2UAT → live
live
Week 2 — hero frames & numbers Day 70 — customer service live Day 77 — try-on live Signed early October, live before year end — 2026 budget; run starts 2027.
Both dates assume the 10-system scope and no storefront rebuild — the two unknowns on the next slide can move them, and they shift week for week with the signature.
What it costs

Two builds, two unknowns — a high-level price.

Use Case 01 · AI Immersive Try-On
Build camera, fit, scene studio, motion, look pageEUR 37,500
+ Website integration placeholder20% · EUR 7,500
Use case 01 total incl. integration placeholderEUR 45,000
Monthly maintenance incl. 2,000 renders a monthEUR 1,400 / mo
More in bulk: each 1,000 renders≈ +EUR 500 / mo
Use Case 02 · E2E AI Customer Service
Build map & measure, shoppers, trade deskEUR 24,000
Use case 02 total incl. 10 connected systemsEUR 24,000
Monthly maintenance incl. 500 conversations a monthEUR 900 / mo
More in bulk: each 500 conversations≈ +EUR 100 / mo
If more systems: each further block of 10+15% · EUR 3,600
Total build
EUR 69,000
incl. EUR 7,500 placeholder
Total maintenance
EUR 2,300 / mo
EUR 27,600 a year
Unknown · placeholder
How your storefront is built
We haven’t seen the site’s structure yet, so we don’t know whether we integrate into it or hand over the code. Placeholder: +20% of the try-on build, EUR 7,500, assuming a standard storefront. A bespoke or headless front end is re-quoted upward once we’ve looked.
Unknown · placeholder
How many systems to connect
Shipping providers, carriers, the platform, the ticket system. The price covers 10 external APIs. Every further block of 10, or part of it: +15% of the customer-service build · EUR 3,600.
up to 10 APIsEUR 24,000included
up to 20 APIsEUR 27,600+15% · EUR 3,600
up to 30 APIsEUR 31,200+30% · EUR 7,200
Indicative. Both placeholders are re-quoted at kickoff — once we see the storefront and list the systems. Maintenance is reviewed then too.
Maintenance covers hosting, monitoring, updates, support and the AI usage for the volumes included above. Past them, volume is bought in bulk — every render and every conversation is a model call. Bulk prices are an order of magnitude, ≈ EUR 0.50 a render and ≈ EUR 0.20 a conversation, confirmed from week 1.
Public support to check with your advisor: SIMEST 10% grant up to EUR 100k, open now · Lombardy SI4.0 50% up to EUR 30k, closes 25 Sep.
Brainstorm · not a proposal yet

Your turn: what did we get wrong?

01
Sales or operations first?
02
Does the try-on match the render you imagined?
03
Which audience hurts most today?
04
Would you A/B test the 15-minute offer?
05
What’s missing?
Retrosuperfuture × defil.ai
The numbers are indicative — the formal proposal follows this conversation.
What we’ll need from you, to start

A short list.

For Use Case 01
Hero frames + packshots
Packshots already on Shopify; a shortlist to tune.
packshots→30 hero frames→tuned in week 2
For Use Case 01
Moodboard & palette
Scenes, outfits, places; what to avoid.
For both
A few hours
Your colleague, creative team, platform builder.
For Use Case 02
The inbox, three months
Tickets, chats, emails — messy is fine.
For both
An NDA and SA
To see how you really work.
For Use Case 02
Rules + read access
policieswarrantytrade terms
ordersshipmentsstock
The value, side by side

What each build is worth — numbers we stand behind.

The try-on
frame sales · UK independent
eyewear sales online · Euromonitor 2026
your variants end in “find a stockist”
Every shared look is media you didn’t pay for expected effect
UK independent optician case (2026) · Euromonitor eyewear e-commerce 2026 · retrosuperfuture.com variant availability, read 15 Sep 2026
Before the next hire
A second hire
EUR 34–42k / year
The assistant · from 2027
EUR 10,800 / year
The build · 2026
EUR 24,000≈ two thirds of a hire
CCNL Commercio 3°/4° livello tables 2026 · employer-cost calculators (INPS ~24–28%, TFR 6.9%, INAIL) · Talent.com Milan · Italian recruiting benchmarks 2026
Both builds plus a year of run ≈ two and a half hires. From year two, EUR 27,600 a year — less than one hire.
Get started

Three steps to kickoff.

Step 01

Green light

A short call to walk through the plan and shape both builds.

Step 02

NDA & SA

NDA and SA with the statement of work; you sign — a week.

Step 03

Kickoff & Week 0

Hero frames, moodboard, inbox export — both builds start; live by year end.

Retrosuperfuture × defil.ai

Thank you.

Make the shop window disruptive again — and take the routine off the one desk that answers everyone.

defil.ai
←→ navigate · space reveal