Shalini Verma in a limestone studio, a laptop on her knees, looking towards the window

Cover story · No. 01 · Toronto · Autumn 2026

One person. The whole loop.

This issue8Agents in production~19%AI-answer citations14Meetings in 6 weeks

AI Marketing Engineer. I replace manual marketing work with agents and automations, then prove it in pipeline, revenue, CAC and hours saved.

Start a conversation Watch the premiere Résumé

Currently leading AI marketing at Imagera AI. Open to a select full-time role leading AI marketing or GTM engineering.

Magazine cover: Shalini Verma, issue No. 01

The issue

8Agents in production · Claude Code agents with golden-set evals and a human review gatep. 12
~19%AI-answer citations · Up from ~8%, tracked through to signupsp. 13
14Meetings in 6 weeks · From 200 target accounts: 72% data match, 9% replyp. 16

The cover: GPT Image 2, made from three of her real photos. Its brief is on p. 40.

In this issue

In this issue. Seven systems, one way of working.

I find the manual work that slows a marketing team down. Then I build the system that does it, and prove it in revenue. Seven systems from the record, the story behind them, the craft behind every picture, the studio and the method. Read in any order.

p. 02The premiereThe title sequence, full screen. One person, one loop.
p. 04Editor’s letterWhat teams bring me in for, and where each is proven.
p. 06The storySeven years. One question. Why do people act, and how do you build the system that makes it happen again?
p. 12FeaturesSeven systems: goal, what was broken, what I built, what changed.
p. 26The profileIn her studio: psychology, jewellery design, luxury brand, AI — and painting.
p. 32The campaignThe one-sheet and seven posters, one per system.
p. 40The making of this issuePick any picture and read the brief that made it.
p. 60Field notesWhat operators learned about AI marketing this year.
p. 64The IndexThe market, in numbers — sources linked.
p. 70The psychology columnNine hooks I use, with the research.
p. 78MastheadThe résumé, in full.
p. 86MethodHow I begin every engagement.
BackBack coverIntroductions, by appointment.

Shalini outdoors in sunglasses and a black blazer, holding a tablet — Shalini Verma · Toronto, Canada. Real photo.

Shot 01 · Watch · Crash zoom in — AI film

Shalini Verma presents

One person. One loop.

AI Marketing Engineer. I replace manual marketing work with agents and automations, then prove it in pipeline, revenue, CAC and hours saved.

8Agents in production
~19%AI-answer citations
14Meetings in 6 weeks

Start a conversation Watch the edit Résumé

Currently leading AI marketing at Imagera AI. Open to a select full-time role leading AI marketing or GTM engineering.

p. 04

Editor’s letter

What teams bring me in for. And where each one is proven.

01
Marketing output that scales without new headcount
8 agents with evals: drafting 45 → 8 minutes
02
To show up when buyers ask AI
AI-answer citations ~8% → ~19%, tracked to signups
03
Budget decisions the CFO will sign
Spend tied to revenue, LTV and payback; executive ROI reviews
04
Pipeline without an SDR army
200 target accounts → 14 meetings in 6 weeks
05
Retention that proves its own lift
+22% vs a 10% holdout; journeys built to cut churn
06
Creative that beats the control
CPA −31%; batches 3 hours → 25 minutes
07
Someone who ships, not someone who briefs
Agents, automations and this site, built by her

p. 06

The story

Seven years. One question.

Why do people act, and how do you build the system that makes it happen again?

AI illustration
2018
The psychologist
Continua Kids
Why people act.
Assessments and parent counselling: the behavioural frameworks behind every hook since.
AI illustration
2019
The teacher
JIMS Rohini
Turn a craft into a method others can run.
~180 students · project grades ~18% higher
AI illustration
2021
The jeweller
Tola and Carat Jewellers
Desire, then proof, then urgency.
Meta click-through ~0.9% → ~2.4% · CPC −38%
AI illustration
2023
Enterprise
Nestlé
Systems at scale, with brand and legal intact.
CRM click-through +35% · consented signups +40%
AI illustration
2024
Prove the lift
Klaviyo · Customer.io
Every journey against a control.
CPA −31% at Klaviyo · +22% vs a 10% holdout at Customer.io
AI illustration
2024
Signal to pipeline
Attio
From a cold list to a booked meeting.
200 target accounts · 14 meetings in 6 weeks
AI illustration
2025
The agent era
Imagera AI
Agents that work, and a gate that keeps them honest.
Drafting 45 → 8 minutes · AI-answer citations ~8% → ~19%

p. 12

Features

Seven systems. Goal, what was broken, what I built, what changed.

Case 01Imagera AICanada · remote

Eight Agents, One Review Gate

Drafting 45 → 8 minutes · first-pass approval 55% → 88% · creative output 4×.

Goal: Launch-grade marketing output every day, without adding headcount or letting quality drift.
What was broken: A draft took 45 minutes and only 55% passed first review. Output was capped by hand work.
What I built: Eight marketing agents in Claude Code: Agent Skills hold the brand rules, MCP connects Meta Ads, GA4 and Ahrefs, a golden set scored by an LLM judge checks every change, and a person approves before anything ships. Each agent has an owner, an eval set and a rollback path; a weekly review keeps first-pass quality above 85%.

Brief
Brand skill
Agent draft
Golden-set eval
Human gate
Publish

Claude Code
Agent Skills
MCP
LLM-as-judge
GitHub

Open the project

Case 02Imagera AICanada · remote

Cited by the Answer Engines

AI-answer citation share ~8% → ~19%, tracked through to signups.

Goal: Be the brand AI assistants name when buyers ask.
What was broken: The brand appeared in about 8% of tracked AI answers, and no line ran from a citation to a signup.
What I built: A weekly citation check across the AI answer engines, structured data (schema.org) and fixes to the pages that earn the mention, with every citation tracked through to signups.

Buyer questions
Weekly check
Gap list
Page fixes
Structured data
Signups

Python
Profound
Ahrefs
schema.org
GA4

Open the project

Case 03Imagera AICanada · remote

Where the Budget Actually Pays

Last-click arguments → a reallocation plan with ROI ranges, reviewed with executives.

Goal: Know which channel earns its keep, in revenue rather than clicks.
What was broken: No clear path from spend to sale. Budget decisions were last-click arguments.
What I built: Spend-to-revenue measurement (GA4, Google Tag Manager, Meta Pixel, CAPI, SQL, Python) that ties each signup to revenue, LTV and payback, and a Meridian incrementality read with a geo holdout. It goes to executive ROI reviews with a reallocation plan.

Track
Attribute
Model
ROI range
Executive review
Reallocate

GA4
Google Tag Manager
CAPI
SQL
Python
Meridian

Open the project

Case 04Imagera AICanada · remote

Always On

Always-on publishing, consistent daily revenue and ~10 hours a week back to the team.

Goal: Publish and run ads every day without a person babysitting the queue.
What was broken: Social and ads ran by hand from a posting queue.
What I built: An always-on calendar in n8n with webhooks into Meta, LinkedIn, TikTok, YouTube and Google Ads: the team approves, then it publishes. Freelance editors only for the final human polish.

Calendar
Draft
Approve
Publish
Measure
Reuse

n8n
Webhooks
Meta
LinkedIn
TikTok
YouTube
Google Ads

Open the project

Case 05AttioRemote

Cold Lists, Warm Meetings

200 target accounts · 72% data match · 9% reply · 14 meetings in 6 weeks.

Goal: Pipeline from target accounts, without sales chasing lists by hand.
What was broken: Sales chased cold lists by hand.
What I built: A proprietary outbound workflow: target list → enrichment (Clay, Apollo) → n8n → a personal draft → the owner approves in Slack → send. SPF, DKIM and DMARC set up; clean data and attribution on every send.

Target list
Enrich
Draft
Slack approval
Send
Meeting

Clay
Apollo
n8n
Attio
Slack
SPF · DKIM · DMARC

Open the project

Case 06Customer.ioRemote

Journeys With a Holdout

Nurture days → minutes · engagement +22% vs a 10% holdout · judged on LTV, payback and cohort quality.

Goal: Nurture that responds in minutes and proves its own lift.
What was broken: Nurture waited days, and journeys were judged on clicks.
What I built: Event-triggered onboarding, expansion and win-back journeys (Customer.io, Liquid, SQL, webhooks), built to reduce churn. A 10% holdout never gets the email, so the lift is real; new segments and languages are a setting. AI-personalised copy stays behind a human review gate.

Event
Segment
Journey
Holdout
Lift
LTV read

Customer.io
Liquid
SQL
Webhooks
Holdouts

Open the project

Case 07KlaviyoRemote

Ads That Beat the Control

3 hours → 25 minutes per batch · CPA −31% vs control.

Goal: Test more creative, faster, and keep only what wins.
What was broken: A batch of ad and email creative took three hours, with no shared view from click to revenue.
What I built: Variants drafted with Claude and Figma, A/B tests against a control in Meta and Google Ads (Performance Max, Smart Bidding), a fixed experiment log, and one view from ad click to Klaviyo signup and revenue (GA4, Google Tag Manager, Meta Pixel, CAPI).

Offer
Variants
Test vs control
Log
Keep the winner
Reuse

Klaviyo
Claude
Figma
Meta Ads
Google Ads
GA4

Open the project

p. 26

The profile · In her studio

I paint on weekends. It is the same muscle I use at work.

Shalini painting a lakeside village at the easel

Shalini in her studio beside a coastal oil painting

Shalini holding up a finished painting of a sea cliff and flowers

Shalini at her table with a watercolour sketch, smiling

Real photos: her own weekend paintings, coastal oils in the Riviera tradition.

I started in psychology, assessing how children learn and why people act. Every hook I write still starts there.

Then luxury. As brand and creative lead for a jeweller, psychology-led creative took Meta click-through from about 0.9% to 2.4%.

Then scale: CRM and content for Nestlé, then Klaviyo, Customer.io and Attio. Each time, a system replaced the hand work and proved its lift.

Now I lead AI marketing at Imagera AI, where eight agents with evals and a human review gate do in eight minutes what took forty-five.

I paint on weekends. It is the same muscle: look closely, then commit to the stroke.

p. 32

The campaign

The campaign. The one-sheet and seven posters, one per system.

AI illustrationOne-sheet — ONE PERSON. ONE LOOP.

AI illustrationPoster — The Content Engine

AI illustrationPoster — Cited by the Answer Engines

AI illustrationPoster — Where the Budget Actually Pays

AI illustrationPoster — Cold Lists, Warm Meetings

AI illustrationPoster — Journeys With a Holdout

AI illustrationPoster — Ads That Beat the Control

p. 40

The making of this issue

Every image here started as a written brief. Pick one and read it.

I write image prompts the way I write a creative brief: task, subject, light, texture, exact text, constraints. Prompted by Shalini · run through the fal.ai API from a Claude Code script. Copy any of them.

Action still · identity from 3 of her real photos

The brief

VFX 1 — Explosion: the deck is dead

Task: A cinematic action-film still, 16:9, the hero walking away from a blast.
Subject: Use the woman in the reference photos as the subject. Keep her exact face, eye shape, brows, nose, lips, skin tone, long dark centre-parted hair and pearl stud earrings. She must be clearly recognisable as the same person. She walks towards the camera in a long black coat over a black suit, calm and unhurried, eyes just past the lens.
VFX: Behind her, in a vast concrete warehouse, a towering stack of printed slide decks and weekly reports bursts outward in a storm of paper: thousands of pages tumbling through the air, a warm flash of light, sparks and embers drifting past her.
Camera & light: Low angle, 35mm anamorphic, backlit by the blast so her silhouette is rimmed in gold, soft fill on her face, light haze, a subtle horizontal lens flare.
Texture: Photorealistic. Real skin texture: visible pores, fine vellus hair, natural variation in tone, no retouching, never plastic. Shot on 35mm film: visible fine organic grain, gentle halation around highlights, true-to-life colour, razor-sharp focus where it matters, micro-detail that holds at full size.
Constraints: No readable text on any page. No logos, no watermark, no signature, no extra text, no extra people.

GPT Image 2 · Edit · her photos as identity reference2560x1440quality high

Prompted by Shalini · run through the fal.ai API from a Claude Code script

Two films from the issue

Title sequence

Kinetic type and the six storyboard frames on one GSAP timeline, rendered frame by frame from code.

The loop, explained

Watch → Build → Convert → Compound as one animated diagram; each lap compounds.

p. 60

Field notes

What operators learned about AI marketing this year. And my read on each.

Full AI: 0.9% CTR, ROAS 2.1. AI-assisted: 2.3% CTR, ROAS 4.7. Human: 2.2% CTR, ROAS 4.8.
Agency test · $92k ad spend · 150 adsr/advertising · 2026-04-02paraphrased
My read AI-assisted matched human results for $680 against $8,000 across 75 ads. AI is a speed layer, not a replacement: a person still owns the hook and the edit.
Read the thread
“creative budget: $8k/month → $2.5k/month”
Performance marketer testing AI video adsr/AskMarketing · 2026-04-08
My read Test 30–40 angles with AI, kill the losers in three days, remake the winners with real creators. Testing gets cheap; taste decides what scales.
Read the thread
“the hybrid format outperformed the full AI version by 31 percent on click-through rate and by 24 percent on cost per purchase.”
Brand running AI-hook ads on Meta, six-week testr/AI_UGC_Marketing · 2026-03-27
My read Let AI iterate the first four seconds, then show the real thing. Hooks move fast; trust is earned with real footage.
Read the thread
“around month 3: quality drift. You stop reading every output … and that is exactly when something subtly wrong ships.”
Marketer, under a zero-ad-spend growth story with 643 upvotesr/ClaudeAI · 2026-04-29
My read Velocity is the easy part. Every batch from my content engine passes a review sheet before anything ships.
Read the thread
“we cut low-effort data warehouse tickets by 73%.”
Analytics team with a five-minute Slack botr/AnalyticsAutomation · 2026-03-04
My read The best automation is small and lives where the team already works. That is why my outbound drafts wait for a yes in Slack before anything sends.
Read the thread
“any ad featuring an ai generated person has to be labeled as a synthetic performer”
Advertiser on the 2026 labelling shiftr/artificial · 2026-08-14
My read Label AI work plainly, as every image on this site is, and compete on the idea instead of on fooling the feed.
Read the thread

p. 64

The Index

The market, in numbers. Sources linked.

live marketing job posts that mention AI list designing or shipping AI agents and automated workflows as a requirement; Claude is named in 26 of them: 45 of 60
MarTech, review of 60 live postings (Margaret Lee) · 2026-09-28
revenue growth for “future-built” AI companies vs laggards — plus 2.3× shareholder return and 2.8× EBITDA growth: 2.4×
BCG, Applied AI Index 2026 (1,330 leaders) · 2026-09-30
average wage premium for workers with AI skills, up from 57% a year earlier: 62%
PwC, 2026 Global AI Jobs Barometer · 2026-06-18
faster growth for jobs needing AI skills (69%) than for the job market overall (9%): ~8×
PwC, 2026 Global AI Jobs Barometer · 2026-06-18
“Growth marketing is the only marketing role that’s actually up year-on-year.”: Only one
Growth Unhinged (Kyle Poyar), Sumble job data, Q1 2026 · 2026-05-20
sales growth for B2B firms that kept marketing through the 1981–82 recession vs those that cut (600 firms, 1980–85): 275% vs 19%
McGraw-Hill Research, via Ad Age · classic study
Wix bought Base44 — built by essentially one founder in six months — for $80M in cash: $80M
Yahoo Finance; Lenny’s Newsletter · 2025-06-18
from Lovable’s first $1M to $100M ARR: 8 months
Lovable (company blog) · 2025-07-23

p. 70

The psychology column

Nine hooks I use, and the research behind each. Turn a card.

p. 78

Masthead

The record. Résumé, in full.

Jan 2025 – now
Imagera AI
AI Marketing Lead · Contract, full-time · Canada · remote
A growth-stage AI product. Led a pod of 5.
- Built 8 production marketing agents (Claude Code, Agent Skills, MCP) with golden-set evals and a human review gate: drafting 45 → 8 minutes, first-pass approval 55% → 88%, creative output 4×.
- Automated social and paid publishing with n8n across Meta, LinkedIn, TikTok, YouTube and Google Ads, for consistent daily revenue.
- Raised AI-answer citation share (AEO / GEO) from ~8% to ~19% with structured data, tracked through to signups.
- Built spend-to-revenue measurement tying signups to revenue, LTV and payback; a Meridian incrementality read replaced last-click.
- Lead executive meetings and ROI showcases: spend, revenue, payback and hours saved per system, with budget reallocation recommendations.
- Every agent has an owner, an eval set and a rollback path; a weekly golden-set review keeps first-pass quality above 85%.
Sep 2024 – Jan 2025
Attio
GTM & Marketing Automation Lead · Contract, part-time · Remote
Led 3–4 contractors per build.
- Built an outbound engine (Clay, Apollo, n8n, Slack approval, SPF / DKIM / DMARC): 200 target accounts, 72% data match, 9% reply rate, 14 meetings in 6 weeks.
- Ran stakeholder meetings with sales and support leaders to scope each build and present pipeline and ROI results.
- Ran 3–5 campaigns a month; new segments and languages became settings, not rebuilds.
May – Dec 2024
Customer.io
Lifecycle & Marketing Automation Lead · Contract, part-time · Remote
Worked with RevOps and two specialists.
- Rebuilt event-triggered nurture (Customer.io, Liquid, SQL, webhooks): response time days → minutes; email engagement +22% vs a 10% holdout.
- Built onboarding, expansion and win-back journeys to reduce churn and lift retention and net revenue.
- Moved journey reporting from clicks to LTV, payback and cohort quality; AI-personalised copy kept behind human review.
Jan – Apr 2024
Klaviyo
Growth Marketing Lead · Contract, part-time · Remote
A creative and SDR pod of 4.
- Cut ad and email creative batches from 3 hours to 25 minutes (Claude, Figma); winning variants cut CPA 31% and improved ROAS and CAC vs control.
- Built ad-click-to-revenue tracking (GA4, Google Tag Manager, Meta Pixel, CAPI, Klaviyo events) and an experiment log for Performance Max and Smart Bidding tests.
Mar – Dec 2023
Nestlé
Digital Marketing Lead, CRM & Content · Contract · Remote
Enterprise CPG. Led an agency and freelance pod of 6–8.
- Lifted CRM click-through 35% and consented signups 40% on key campaigns (Salesforce Marketing Cloud, SQL segmentation).
- Cut campaign asset production from ~10 to ~4 days with a brief-approve-publish workflow, keeping brand and legal review.
- Presented campaign ROI to brand and marketing leadership to guide CRM and media budget decisions.
Jul 2021 – Feb 2023
Tola and Carat Jewellers
Brand & Creative Lead · Full-time · Delhi, India · on-site
Luxury retail. A team of 5.
- Owned brand, Meta and Instagram ads, organic social and in-store creative; hired and briefed freelance photographers, editors and designers.
- Psychology-led ad creative: Meta click-through ~0.9% → ~2.4%, CPC −38%, save and share rate ~2.1×.
- Social-attributed walk-ins ~+45% in festive peaks; average ticket on campaign pieces +22%.
- Creative turnaround ~7 days → ~48 hours; 3× asset volume with the same headcount; engagement +60%.
Feb 2019 – Aug 2021
JIMS Rohini
Guest Assistant Professor · Part-time
- Taught design, manufacturing, quality and merchandising to ~180 students; a new curriculum lifted project grades ~18% and cut incomplete submissions ~40%; mentored 25+ portfolios into placements.
2016 – 2019
Earlier
Psychology, merchandising and design
- Assistant Psychologist, Continua Kids · Jewellery Merchandiser & Designer, Viney Gems and Niaj by Shradha · Design Intern, Woven Gold · Marketing & Design Intern, boutique jewellery brands.

Toolkit

AI & agents

Claude Code
Cursor
Agent Skills
MCP
LLM evals
Golden sets
LLM-as-judge
ChatGPT
Midjourney
Fal.ai

SEO · GEO · AEO

SEO
GEO
AEO
AI-answer citations
Structured data
Profound
Ahrefs

Automation & GTM

n8n
Gumloop
Clay
Apollo
APIs
Webhooks
HubSpot
Salesforce
Attio
SPF · DKIM · DMARC

Lifecycle & CRM

Customer.io
Klaviyo
Braze
Iterable
Salesforce Marketing Cloud
Liquid

Paid & social

Meta Ads
Google Ads
Performance Max
Smart Bidding
LinkedIn
TikTok
YouTube
Figma

Data & measurement

SQL
Python
JavaScript
GA4
Google Tag Manager
Meta Pixel
CAPI
A/B testing
Holdouts
Incrementality
Meridian MMM

Business

Revenue growth
Pipeline
ROI
ROAS
CAC
LTV
Payback
Churn & retention
Unit economics
GTM strategy

Leadership

Executive ROI reviews
Stakeholder management
Budget reallocation
Leading freelance pods

Education & certifications

B.A. Psychology — Indira Gandhi National Open University (IGNOU), 2015 – 2018
Master’s-level coursework in psychological sciences — IGNOU
Diploma in Jewellery Designing — Indian Institute of Gems & Jewellery, 2015 – 2016

Certifications

Claude Code in ActionAnthropic Academy2025
AI-Powered Performance AdsGoogle Skillshop2024
Digital MarketingHubSpot Academy · including AEO and AI for Marketers2023

Claude CodeCursorAgent SkillsMCPLLM evalsGolden setsLLM-as-judgeChatGPTMidjourneyFal.aiSEOGEOAEOAI-answer citationsStructured dataProfoundAhrefsn8nGumloopClayApolloAPIsWebhooksHubSpotSalesforceAttioSPF · DKIM · DMARCCustomer.ioKlaviyoBrazeIterableSalesforce Marketing CloudLiquidMeta AdsGoogle AdsPerformance MaxSmart BiddingLinkedInTikTokYouTubeFigmaSQLPythonJavaScriptGA4Google Tag ManagerMeta PixelCAPIA/B testingHoldoutsIncrementalityMeridian MMMRevenue growthPipelineROIROASCACLTVPaybackChurn & retentionUnit economicsGTM strategyExecutive ROI reviewsStakeholder managementBudget reallocationLeading freelance podsClaude CodeCursorAgent SkillsMCPLLM evalsGolden setsLLM-as-judgeChatGPTMidjourneyFal.aiSEOGEOAEOAI-answer citationsStructured dataProfoundAhrefsn8nGumloopClayApolloAPIsWebhooksHubSpotSalesforceAttioSPF · DKIM · DMARCCustomer.ioKlaviyoBrazeIterableSalesforce Marketing CloudLiquidMeta AdsGoogle AdsPerformance MaxSmart BiddingLinkedInTikTokYouTubeFigmaSQLPythonJavaScriptGA4Google Tag ManagerMeta PixelCAPIA/B testingHoldoutsIncrementalityMeridian MMMRevenue growthPipelineROIROASCACLTVPaybackChurn & retentionUnit economicsGTM strategyExecutive ROI reviewsStakeholder managementBudget reallocationLeading freelance pods

p. 86

Method

How I begin. The first ninety days of any engagement.

Week 1Listen and instrument
Audit tracking, funnels and the manual work. Find where time and money leak.
Day 30First proof
Spend tied to revenue, one automation live behind a human gate, and three bets ranked by ROI.
Day 60First agents
Agents with golden-set evals on the highest-volume work, and a holdout on the first journey.
Day 90Compounding
An executive ROI review, a test cadence and a written playbook the team runs without me in every loop.

Shalini Verma: her name set huge behind her portrait

Back cover · Introductions

Introductions. By appointment.

Currently leading AI marketing at Imagera AI.Open to a select full-time role leading AI marketing or GTM engineering.Advisory and contract work by arrangement.

What would we explore?Full-time leadership roleContract engagementAdvisorySpeaking or workshop

Start a conversation Résumé (PDF)LinkedIn

Roles considered: AI Marketing Engineer · Lead · Head of Growth · AI products · GTM Engineering Lead

shaliniworks24@gmail.com · Toronto, Canada · Remote or hybrid across Canada and the US

One person. The whole loop.

In this issue. Seven systems, one way of working.

One person. One loop.

The psychologist

The teacher

The jeweller

Enterprise

Prove the lift

Signal to pipeline

The agent era

Eight Agents, One Review Gate

Cited by the Answer Engines

Where the Budget Actually Pays

Always On

Cold Lists, Warm Meetings

Journeys With a Holdout

Ads That Beat the Control

VFX 1 — Explosion: the deck is dead

Two films from the issue

Title sequence

The loop, explained

Imagera AI

Attio

Customer.io

Klaviyo

Nestlé

Tola and Carat Jewellers

JIMS Rohini

Earlier

Toolkit

AI & agents

SEO · GEO · AEO

Automation & GTM

Lifecycle & CRM

Paid & social

Data & measurement

Business

Leadership

Education & certifications

Introductions. By appointment.