MAELIFY // TCF
Evaluation Response

An AI-native operating
model for TCF’s
e-commerce line.

Particula, torn down with live data. A service line designed as a P&L. A first 90 days.

JOSE GALIANO · MAELIFY · SHOPIFY PLUS ARCHITECT
START HERE

In my own words.

The reasoning behind everything that follows.

▶
Intro walkthrough

Loading…

00 / APPROACH

AI is not a service line.
It is the operating system.

EVIDENCE

Built agentically

Their live catalog, Lighthouse, Klaviyo sends, ad libraries. Hours, not days.

THE MODEL

Yerevan × AI

A cost and quality frontier a Western agency cannot reach.

THE WEDGE

Retention & CRO first

Highest margin, lowest risk, and where these brands actually bleed.

01 / TEARDOWN

Particula.

Good product. Real paid engine. The layer in between leaks.

64
Published SKUs
31%
Sold out, still listed
18.3s
Homepage LCP, mobile
#1 & #2
Sellers = HP Wizard boards
01a / THE FUNNEL

Walked, end to end.

HOMEPAGE
26 MB, 18 seconds on mobile
GOCHESS PDP · $380
Three claims that contradict each other
GOCUBE PDP · $60
Shop Pay and Prime working hard
CART
A timer that is not real
CHECKOUT
Low friction, then import VAT
Low Stock badge
In stock, ships now
ONE PDP, THREE PROMISES: “Low Stock” above “In stock, Ships now”. A flat 1 Year Warranty beside a warranty conditioned on a paid membership.
Working well
  • Shop Pay, Buy with Prime, every wallet
  • 190 to 530 reviews per hero product
  • Licensed IP that competitors cannot copy
Leaking
  • The #1 and #2 sellers are missing from their own menu
  • Cross-sell app installed, never deployed
  • A $1 “Test” product live in the catalog
01b / WHAT IS COMING

I queried their store
as an AI agent would.

INFRASTRUCTURE
A

Custom llms.txt, live MCP endpoint, UCP 2026, product schema.

WHAT THE AGENT GETS
F

Test dummies, ahead of the products.

“BEST SMART GIFT FOR A CHESS PLAYER”
1  Test · $1, empty description
2  GoChess Wizard Robotic · $0
3  GDN1ACCS-PACK · internal SKU
…
7  GoChess Wizard Lite · their best seller

They built the shop window for AI agents, then filled it with test dummies.

The blind spot behind it

About 70% of AI-referred traffic lands in GA4 as “direct.” The ROAS signal every paid decision leans on is already wrong. Fix: server-side tracking.

01c / LIFECYCLE

Their Klaviyo program,
reverse-engineered.

13:39
Welcome. WIZARD10, $10 flat
13:47
A second welcome. MOVE10, 10%
13:57
Abandoned. Extra5, $5
14:07
Abandoned again. No code
02:07 · 2AM HIS TIME
Wizard flow escalates. ENCHANT5, 5%
09:58
Generic flow flatlines. The same $5
02:09 · TUE, 2AM AGAIN
Wizard hits 10%. ENCHANT10, $38. “This final offer expires”
08:59 · SEVEN HOURS LATER
A second “final offer”. The same $5

FOUR EMAILS IN 28 MINUTES. SENDS ANCHORED TO US HOURS. NO SMS EVER CAME.
THE $5 CODE ARRIVED THREE TIMES IN 43 HOURS, EACH WITH A FRESH COUNTDOWN.

The headline finding

Two final offers. Seven hours apart. Thirty three dollars apart.

Tue 02:09
-$38
Tue 08:59
-$5

Same cart, same morning, both subject-lined the final offer. The later one is worth $33 less.

Five codes in 37 hours: $5, $10, $19, $38, $38. None combinable. The premium flow climbed ($0 → $19 → $38) while the generic flow never moved off $5. Same person, same cart, neither flow able to see the other.

Cart still reserved
CAPTURED THE NEXT MORNING: the cart still claims “reserved for 15 minutes” more than twenty hours later.

Sophisticated machinery. Broken logic underneath.

01d / PRIORITIES

Three fixes.

1
High impact · low effort

Rebuild the lifecycle logic

Escalate the ladder, suppress across flows, segment with Klaviyo AI. Configuration, not engineering.

2
High impact · medium

Recover paid ROAS, and see it

Cut 26 MB, deploy the cross-sell they already pay for, fix attribution server-side.

3
Compounding

Make the store agent-ready

Clean the feed before AI discovery matters. Cheap now, expensive later.

What I would want that I cannot see

Klaviyo revenue by flow. Checkout drop-off. Per-SKU margin. Membership churn. Server-side AI referrals. Each one turns a ranked guess into a costed decision.

02 / SERVICE LINE

Designed as a P&L.

Offer
  • AI-augmented Retention & CRO
  • Agentic Commerce Readiness
  • AI Creative Studio
  • Full Growth Management
Refuse
  • Headless and custom builds
  • Marketplace management
  • Brand identity work
  • Autonomous customer-facing AI

A new line earns the right to expand by being narrow first.

02a / THE POD

Human judgment,
agent throughput.

CRO senior
+ Coframe, Webflow Optimize
Lifecycle mid
+ Klaviyo Segments AI
Creative mid
+ Arcads, Creatify, HeyGen
Developer 0.5
+ AI codegen
Account junior
+ always-on ops agent

ONE SENIOR ANCHORS QUALITY. AGENT DRAFTS, HUMAN SHIPS. EVERY CUSTOMER-FACING OUTPUT PASSES A HUMAN GATE.

02b / THE MODEL

Touch trades against leverage.

Retention & CROFull GrowthPaid Media
LeadProductizedHigh touchLight
Pod4.27–81.6
Brands / pod5–63–46–8
Brands / head1.30.54.0
Retainer$5k$12–18k$3–4k
Gross margin68%58%70%
Net EBIT24%19%24%

BLENDED $1,600 / HEAD / MONTH, IE MODEL. CAPACITY IS QUALITY-CONSTRAINED, NOT COST-CONSTRAINED.

02c / THE MOAT

Not a margin trick.

Retainer
$5.0k
Pod labour
-$1.5k
AI + QA
-$0.6k
Gross margin
68%
Ramp, CAC, G&A
-34%
Net EBIT
24%

24% EBIT, honestly derived, beats an 85% gross margin nobody believes.

  • The constraint is quality, not cost
  • So you price to value, never cost-plus
  • And reinvest margin into QA and attribution
02d / MECHANICS

What makes it real.

RAMP

Paid onboarding

A $20k accelerator absorbs the build-heavy first weeks.

FIT

ICP gating

CRO does nothing below $500k a month. Know when not to sell.

PRICING

Base + performance

A share of attributed revenue, tied to real KPIs.

RISK

Where AI breaks

Hallucination, attribution loss, key-person risk. All hedged.

02e / RECOMMENDATION

Lead with Retention & CRO.

  • Highest leverage, lowest delivery risk
  • Where post-crowdfunding brands actually bleed
  • Leads from my depth, not the paid seat
  • The wedge into full management
STAGE 1
One pod, two pilot brands
STAGE 2
Fill the pod, systematize
STAGE 3
Pod two, senior oversight
03 / FIRST 90 DAYS

Build the org, not the headcount.

BUILD

The AI-native stack

Accelerator SOP, flow and test libraries, ops agents.

HIRE

A specialist pair

Senior lifecycle and CRO together, plus a fractional dev.

OWN

Strategy and the first accounts

I set the bar by doing the work before delegating it.

DELEGATE

Execution into the system

Agents carry monitoring. I move from doing to reviewing.

0–30 · FOUNDATION
  • Offer, pricing, ICP
  • Hire the pair
  • Two pilot brands
30–60 · PROOF
  • First accelerators shipped
  • Server-side attribution live
  • Margin proven
60–90 · SYSTEMIZE
  • QA gates hardened
  • Pod filled
  • Economics to leadership
The through-line

Design the business that employs the specialists.

JOSE GALIANO · MAELIFY · jose.galiano.tms@gmail.com
MAELIFY // PREPARED FOR THE CROWDFUNDING FORMULA // CONFIDENTIAL