AI-assisted marketing / creative & commercial systems

Shelby
Youtsas

I combine AI-assisted creative work with commercial analysis and consumer-brand experience.

This is the online edition of my portfolio for the AI Marketing Specialist role at Gelato. Everything with a slider is live. Start with the Campaign Lab.

Open the Campaign Lab See the work

Mathematics minor: plus 3.5 years of chemical engineering study and a BA in English, Professional Writing
Claude and ChatGPT: across model variants, every working day, for research, modeling, writing and visuals
Meta Ads, Google Ads and SEO: analytics in four roles, with campaigns and ad spend run in three

AI-assisted concept rendering of a Spanish Revival kitchen with olive cabinetry, plaster hood and terracotta floor tile — AI-assisted creative concept / Spanish Revival study

Selected work

8 projects

Work, sorted by what it proves.

01Campaign LabWorking prototype · paid economics, AI briefs, test plans, monitoring
02AI-directed creative developmentStructured prompting · iterative visual direction
03Commercial scenario modelingEight-worksheet toolkit · escalation, discounting, scenarios
04Client decision toolsAI-built calculators · catching the model's mistakes
05Hannah Castille DesignsBrand DNA · five identity directions · content plan
06Listing DiagnosticsProductized service · offer, copy and artwork
07Design culture, translatedPublished essays · seven-question style experience
08A broader point of viewRenovation, retail training, hospitality planning

01 / Application prototype

Built for Gelato

Campaign Lab

A creative and commercial loop: set the economics, generate the creative, plan the test, watch the results.

Pick a product, market, occasion, audience and channel. Each tab builds on the one before it, the way I'd want the work to run in practice: the unit economics decide the budget, the budget decides how big a test can be, and the monitoring rules protect the spend once it's live.

My role: workflow design, AI-assisted implementation, metric definitions, prompt templates and validation. All inputs are illustrative, not Gelato data. Calculations and templates run in your browser, with no live AI or ad-platform connection.

Product

Market

Occasion

Audience

Channel

Daily budget

Cost per click at €4,000 a day

Conversion rate

Average order value

Gross margin

New customers, share of orders

CPC rise each time spend doubles

Models diminishing returns: more budget buys pricier clicks.

ROAS

Revenue ÷ ad spend

Acquisition cost

Ad spend ÷ new customers

Contribution / month

After variable costs and ads

Break-even ROAS

1 ÷ gross margin

Marginal ROAS

Return on the next euro

Profit-max budget

Per day, where marginal ROAS = break-even

Monthly contribution by daily budgetYour budgetProfit-maximizing budget

Show the math

Sample size

Smallest lift worth detecting

Relative lift in conversion rate. 95% confidence, 80% power, two-sided.

Clicks per variant

Days to read

Test spend

Held constant

Decision rules

Named assets

Asset name	Angle	Format

Naming: brand_market_product_occasion_angle_format_version. One variable changes per test: the creative angle.

Rules an agent runs every morning

Pacing tolerance vs plan

Flag CAC above trailing 7-day average by

Flag conversions down, clicks steady, by

Separates a tracking break from a real demand drop.

Acquisition cost, last 14 days

Flags

Market	Day	Rule	What to do

Draft for the team channel

Illustrative data generated in your browser from the Economics tab, with three problems planted to test the rules.

02 / AI-assisted creative

Renovation concept

AI-directed creative development

Concept study with AI-assisted visualizations

AI-assisted concept rendering of a Spanish Revival kitchen with wood tall cabinets, wall ovens and olive lower cabinets

Two views from the same coordinated concept set, developed through iterative direction.

GatherClient references and room constraints
StructureA prompt with materials, light and layout spelled out
ReviewCheck the render against what can actually be built
RefineSpecific visual feedback, one change at a time
DecideConcepts the client can choose from before buying

A repeatable creative process that turns client references and constraints into photorealistic concepts, then refines them through specific visual feedback.

My role

Creative direction, structured prompting, visual review and iterative refinement, using Claude and ChatGPT across research, writing and creative development.

The work

A coordinated set of AI-assisted concept visuals that let the client make decisions before procurement and construction.

Why it matters at Gelato

The same loop scales to ad creative: brief, generate, check against what the product really looks like, refine, ship. The Campaign Lab's brief tab is this process, written down as templates.

03 / Quantitative & commercial

Original model architecture

Commercial scenario modeling

Original spreadsheet architecture and formula excerpts

Commercial Toolkit / NNN lease schedule

Base rent, year 1 Inputs!B4

Annual escalation B5

Discount rate B6

Term in years

Formula structure from the original workbook. Inputs here are illustrative; private client inputs and outputs are omitted.

Linked spreadsheet models that turn editable assumptions into cost, cash-flow and scenario comparisons.

My role

Model structure, spreadsheet logic, scenario comparison and client explanation. Claude and ChatGPT support my current modeling and analysis work.

The work

An eight-worksheet Commercial Toolkit (shared inputs, linked lease, purchase and alternative-return schedules, and a comparison dashboard) plus separate renovation and buyer-decision models.

How it's built

One inputs sheet; every schedule reads from it
Escalation and discounting as explicit, auditable columns
The dashboard only links to results, so nothing is typed twice

04 / AI-built client tools

Real estate & creative-services clients

Client decision tools

Built with AI, checked by hand

Rent cut or repaint?

A fourplex owner could cut rent to stay competitive, or pay a painter and hold rent. Claude built the first model fast. It also made one mistake.

Annual benefit of holding rent

Months to pay back the painter

Illustrative numbers: four units, $1,200 held vs. $1,100 cut, $6,000 paint job, 40% operating-expense ratio, 5% vacancy.

Two pricing models, one business

A creative-services client charged hourly on some projects and a percentage of the budget on others. On a typical project, the hourly model paid about a third as much.

Project budget

Percentage fee

Hourly rate

Hours on the project

Percentage model

Hourly model

Illustrative numbers. The fix was a single rate card: consultation credited toward booking, fixed-fee packages, a percentage fee with a minimum for full-home work, half due at signing.

Decision tools clients can use themselves, published as web pages instead of buried in spreadsheets.

My role

Framing the decision, directing Claude to build the model and visuals, then checking every formula before a client sees it.

The work

A rent-vs-renovate model benchmarked against the S&P 500, with a proposal email and four infographics, including a slope chart of the building's value gap over time
A rent, sell or invest calculator for homeowners, covering Section 121, depreciation recapture, 1031 exchanges and Louisiana closing costs, with a house-price slider and a 4% withdrawal framing
A pricing diagnosis and new rate card for a creative-services client

The catch

The AI applied fixed operating expenses to a marginal rent difference. Operating costs don't change with the rent you charge; only vacancy does. Toggle the first widget to see what that error does to the answer.

05 / Brand & strategy

Client engagement, 2026

Hannah Castille Designs

Delivered strategy and identity directions

Finding the language for an artist-led brand.

An interior designer's point of view, translated into a usable brand strategy that connects positioning and voice with a visual system and a plan for what to publish next.

My role

Brand research, positioning, tone of voice, visual direction, identity exploration and content architecture.

The work

A 14-page Brand DNA document, five identity directions, content pillars revised with the client's feedback, and a tiered social-content and implementation plan priced against local agency rates.

Outcome

The client chose Direction 02, The Canvas, and it was refined into her final mark.

06 / Brand & service strategy

Published offer

Listing Diagnostics

Published service architecture and marketing

A productized service for agents who need a clear explanation of why a listing is losing attention, and what to change first.

My role

Service architecture, offer naming, positioning, sales copy, visual direction and client deliverable design.

The work

A live service range with defined scopes, supporting copy, branded artwork and individual add-ons, starting at a $58 entry offer that includes a seller price-drop script.

The marketing lesson

A low-priced, specific first offer is an acquisition product. It's the same logic as a starter print order: get the first purchase, earn the next one.

Explore the work online

07 / Editorial & digital

youtsas.com

Design culture, translated

Published writing and interactive website content

Published editorial / portfolio excerpt The new
Americana
dream. Culture, home, and the way we live

A point of view people can read and use.

Published writing and a style-discovery experience that connect cultural interests and the way people live with decisions about the home.

My role

Editorial concepts, original writing, brand voice, design interpretation and website experience direction.

The work

Published essays and a seven-question style experience on youtsas.com. Headlines are reset here as excerpts; links lead to the original work.

Why it matters for search

Content people actually read and use is what earns organic traffic, and more and more it's what AI assistants quote when someone asks for a recommendation.

Read the published work

08 / Broader creative range

Selected

A broader point of view.

The Teche renovation

Completed interiors, finish selection, hands-on work, styling and original project storytelling, documented on The Flip Chronicles. Every flip starts as a P&L: purchase price, scope, and an exit price backed by price-per-square-foot comps.

The Maven way

Original retail training materials connecting product knowledge to the customer conversation, written while helping the business grow fourfold.

A growth plan for a downtown inn

I first wrote an acquisition plan for prospective buyers of a five-bedroom short-term rental. When the owners decided to keep it, I rebuilt it as an owner-operator growth plan, built around profit and the guest experience, with a 2027 marketing calendar.

Also available on request

The Union pitch book, the Daylily launch email, event creative and published redesign concepts.

How I work with AI

Four steps, every project

AI as the default layer.

Frame the decision
Name the choice, the inputs and the P&L line it moves before opening a chat.
Let AI take the first pass
Claude and ChatGPT draft the research, model, copy or code. Claude connects to my Gmail, Google Drive, Calendar and Adobe through MCP, so it works from real files.
Check the logic
I read every formula the way my engineering coursework taught me to. That's where the rent-model error got caught.
Ship a working tool
The deliverable is something people can use: a calculator, a page, a plan. This portfolio is one.

At Gelato, I'd point the same loop at budget pacing, search-term mining, anomaly alerts and creative testing, and document each one so the team can reuse it once it proves itself.

Background & measurement experience

2014 to present

Quantitative foundations.

My quantitative background includes a mathematics minor and 3.5 years of chemical engineering study. I pair that with a BA in English focused on Professional Writing from the University of Louisiana at Lafayette, so the analysis comes with a clear explanation.

Current AI workflow

Claude across model variants and ChatGPT support my modeling, research, writing, proposal development and iterative visual work.

Paid media across four roles

I planned, launched and optimized Meta and Google Ads campaigns and managed the spend at Maven, Parish and my own business. At Stuller, I analyzed SEO, Meta and Google Ads performance alongside sales data.

Credentials

Licensed Realtor in Louisiana. Owner-operator of my own LLC.

2022 to present
Shelby Youtsas Real Estate & Design
Owner, Realtor and interior designer
- Run Meta and Google Ads campaigns and budgets for listings and the brand
- Build AI-assisted client tools, brand strategy and marketing plans
2021
Stuller, Inc.
Fine Jewelry Product Manager
- Owned product forecasting: demand, inventory and revenue, three quarters ahead
- SEO, Meta and Google Ads analytics alongside sales data
- Contributed to more than $500K in business growth
2019 to 2020
Parish Brewing Co.
Social Media & Marketing Coordinator
- Planned and ran Meta and Google Ads campaigns and managed the spend
- Grew the social following by 7,000 in seven months
2014 to 2019
Maven Womenswear
Brand Manager
- Planned and ran Meta and Google Ads campaigns and managed the spend
- Email open rates 10–20% above the industry average
- Helped the business grow fourfold

AI

Claude and ChatGPT, across models
MCP integrations
Generative image workflows

Paid & search

Meta Ads
Google Ads
SEO

Analysis

Advanced Excel
Oracle, MicroStrategy
Lightspeed

Creative & CRM

Adobe Photoshop, Lightroom, Premiere Rush
Mailchimp, Flodesk
TikTok, Instagram

For a considered next move

Let's
make it
good.

shelby@youtsas.com

337-849-3995

linkedin.com/in/youtsas

youtsas.com

Lafayette, Louisiana

One more thing

I'm also a Gelato customer, so I know the assortment from the ordering side, including where it has gaps. I'd like to walk you through them.

The portfolio

AI-assisted creative development, commercial modeling, marketing analytics, brand strategy and published content, plus a working Campaign Lab built for this application.

Creative instinct. Analytical precision.

ShelbyYoutsas

Work, sorted by what it proves.

Campaign Lab

AI-directed creative development

My role

The work

Why it matters at Gelato

Commercial scenario modeling

My role

The work

How it's built

Client decision tools

Rent cut or repaint?

Two pricing models, one business

My role

The work

The catch

Hannah Castille Designs

My role

The work

Outcome

Listing Diagnostics

My role

The work

The marketing lesson

Design culture, translated

My role

The work

Why it matters for search

A broader point of view.

The Teche renovation

The Maven way

A growth plan for a downtown inn

Also available on request

AI as the default layer.

Frame the decision

Let AI take the first pass

Check the logic

Ship a working tool

Quantitative foundations.

Current AI workflow

Paid media across four roles

Credentials

AI

Paid & search

Analysis

Creative & CRM

Let'smake itgood.

Shelby Youtsas, Realtor and Design

Shelby
Youtsas

Let's
make it
good.