This $401 Million Company Built by Two People Reveals the New Rules of AI Powered Marketing

In September 2024, Matthew Gallagher launched Medvi, a GLP-1 telehealth startup, from his home in Los Angeles with no employees, no venture capital, and no traditional marketing team. 

By the end of its first full year, Medvi had posted $401 million in sales, served 250,000 customers, and produced a 16.2% net profit margin, nearly triple the margin of Hims & Hers, which employed 2,442 people. Sam Altman’s prediction that AI would produce a one-person billion-dollar company took eighteen months to prove true.

But before we canonise Medvi as the AI marketing gospel, something the headlines missed matters enormously for anyone building a real, durable business. We will get to that. First, the structural picture.

The Great Marketing Reset: What Has Fundamentally Changed

The Medvi story is a data point. What it points to is something that is more of a structural reset of the foundational economics of marketing that has been building for three years and has now arrived all at once.

For the previous thirty years, marketing operated on a stable set of assumptions. 

• Scale required headcount. 
• Reach required budget. 
• Creative quality required agencies. 
• Distribution required relationships. 

Every one of those assumptions was, at some level, a cost barrier and cost barriers are also moats. The company with more people, 

What Nobody Is Telling You About AI and Marketing in 2026

There is a system most marketing organisations have built over the past thirty years that nobody talks about directly, because it is too embedded in how things work to be seen clearly from the inside.

The system is built on a set of assumptions that were entirely reasonable when they were formed. 

• That producing marketing content at scale requires large teams. 
• That reaching a national audience requires substantial budget and agency relationships. 
• That testing creative is an expensive, slow process reserved for major campaigns. 
• That search visibility is a long-term project requiring months of technical work and ongoing investment. 
• That personalising customer communications at scale requires enterprise software and dedicated operations staff.

These assumptions were not wrong. They were accurate descriptions of the cost structure of marketing as it existed. And like all cost structures, they produced an organisational architecture designed to manage them. 

• Teams to handle production. 
• Agencies to handle reach. 
• Budget cycles to govern spending. 
• Approval processes to protect quality. 
• Org charts to coordinate the complexity.

The system worked. For decades, it worked well.

Then the cost structure changed. Not gradually. Not in one area. All at once, across every function that the system had been built to manage.

In 1995, a business owner who wanted to run a national advertising campaign needed a minimum budget of $250,000, an agency, a media buyer, a production team, and publisher relationships that took years to build. The barrier was structural. It was not laziness or lack of ambition that kept most businesses from competing at that level. It was the genuine cost of the infrastructure required.

In 2026, the same reach is available for under $500 a month. Not similar reach. The same reach. Often better targeting. Often faster creative iteration. Often higher margin.

“The danger is not that you have the wrong tools. The danger is that you have built the right organisation for a cost structure that has been retired.”

What the Data Actually Shows

84% of marketing teams are now using AI in at least one workflow. That number sounds like a transformation. 

Then you read the next one: only 17% of those professionals have received comprehensive AI training. The tools have been adopted. The thinking has not changed. The system persists inside a new interface.

Here is the number that should stop everyone in the room: AI-referred web sessions grew 527% year-over-year in 2025. Not 5%. Not 52%. Five hundred and twenty-seven percent. 

The fastest-growing source of web traffic is now AI answer engines:

ChatGPT, Perplexity, Google AI Mode, Claude and fewer than 40% of brands are doing anything to appear in those answers. The rest are investing in search optimisation for a landscape that no longer describes how the majority of information discovery happens.

And from the state of the global workforce: 21% of employees are genuinely engaged in their work. That is not an HR problem. It is a meaning problem. And it costs the global economy $8.9 trillion every year. The teams that will win in this era are not the ones who use AI to move faster inside the old system. They are the ones who use AI to ask what the system should actually be for.

Three Things Most Practitioners Have Not Been Told

First: AI has made authentic human perspective more scarce, not less relevant. 

The explosion of AI-generated content has flooded every channel simultaneously. Almost everything now looks polished, sounds confident, and is forgettable. The content that earns genuine attention that stops the scroll, earns the share, builds the subscriber is the content that could only have come from a specific human with specific experience. The irony of the AI era is that it has created the scarcest thing in the market: genuine, unreproducible point of view.

Second: the biggest AI marketing opportunity is not at the top of the funnel. 

Most conversations about AI marketing are conversations about content production. But the measurable returns from AI are largest inside the funnel: in lead scoring that improves qualification rates by 60%, in onboarding sequences that double Day-30 retention without changing the product, in churn prediction models that identify at-risk customers four weeks before they cancel, in email send-time optimisation that lifts open rates by 35% without a single new word being written. The content story is the visible story. The funnel story is where the money is.

Third: the search game changed while most marketing departments were looking the other way. 

55% of all Google searches now show an AI Overview. These systems do not return a list of blue links. They synthesise an answer and cite sources. 

The brands that appear are the ones with original data, clear structure, and genuine domain authority. 

The brands that do not appear are invisible to the fastest-growing traffic source in the ecosystem. Most of them have not noticed yet because their traditional SEO rankings have not changed. Visibility and traffic have been quietly decoupled.

What This Playbook Is Built to Do

This playbook does not argue that AI will replace marketers. It argues something more specific and more uncomfortable: that marketers who understand what AI is actually for, at each stage of the funnel, in the right sequence, with the right guardrails will produce outcomes that those who do not cannot match. Not because they are smarter. Because they are working with the grain of how the cost structure has changed, rather than against it.

Each of the twelve chapters/sections that follow covers one stage of the marketing system. Each is anchored by a lead expert chosen for their usefulness at that specific stage, a data chart that makes the argument visible, a real tactical example from an operator who has done the work, and live citations to the research behind the numbers.

The sequence is the argument. 

1. Demand intelligence before content creation. 
2. Visibility before distribution. 
3. Workflow before revenue. 
4. Onboarding before retention. 
5. Measurement throughout. 

Most AI marketing advice presents these as parallel options you can adopt in any order. They are not. They are a system. A weakness at any stage compounds downstream. The organisations that understand this are building something durable. 

The ones that do not are using new tools to run an old system faster.

The System at a Glance

Before Chapter One, here is the complete map. Eight stages. One system. Each one built on what came before it.

Stage	Chapter	Lead Expert	AI Leverage Point	Core Metric
Awareness & Visibility	03	Aleyda Solis	Structure content for AI citation (GEO)	55% of searches show AI Overview
Demand Intelligence	02	Rand Fishkin	Research before tool selection	84% use AI; 17% trained
Content Engine	04	Ross Simmonds	One idea → 7 assets via AI	58% higher engagement
Attention & Social	05	Gary Vaynerchuk	Platform-native AI creative iteration	TikTok: +200% follower growth
Workflow Execution	06	Kieran Flanagan	AI agents: research → publish	16 hrs saved/marketer/week
Revenue & Conversion	07	Kipp Bodnar	AI lead scoring + CRM enrichment	1.5× revenue growth vs peers
Onboarding	08	Elena Verna	Personalised time-to-first-value path	Day-30 retention +60%
Retention & Lifecycle	09	Elena Verna	Churn signal detection 3-4 wks early	Expansion revenue +60-90%

The stack is the infrastructure. The moat is what you build with the time the stack gives back to you. Everything that follows is about building the right moat, at the right stage, in the right order.

CHAPTER 01

AI Has Changed the Shape of Marketing

The most important mental model shift of this era is also the simplest: AI is not a tool. It is a new operating layer that sits underneath every function in a modern marketing organisation. Teams that treat it as a productivity add-on will continue to operate on the same model as before, only faster. Teams that understand what has structurally changed will build a different kind of system entirely.

The Three Structural Shifts

First: the marginal cost of content has fallen toward zero. A marketing team that could produce twelve pieces of high-quality content per month in 2021 can now produce sixty or more with the same headcount. The constraint has moved from production capacity to audience attention.

Second: the cost of iteration in paid creative has collapsed. An AI-equipped operator can now generate, test, and iterate on thirty creative variants in the time it used to take to produce three. You no longer need to guess which message or visual resonates.

Third: the search landscape has been restructured from below. AI-referred web sessions grew 527% year-over-year in 2025. The question is no longer just “do I rank on page one of Google?” It is “am I the source that AI systems cite when someone asks the question my content answers?”

“Stop thinking about AI as a tool. Start thinking about it as part of the operating system of modern growth.”

CHAPTER 02

Start With Demand, Not Tools

Rand Fishkin built his reputation by telling marketers things they did not want to hear. His core argument is that most marketing investment is wasted not because of poor execution but because of poor demand intelligence. Teams build content before understanding what their audiences actually care about. They target keywords before verifying that real intent exists behind them.

AI makes this problem worse before it makes it better. A team with strong demand intelligence can use AI to execute faster and at greater scale. A team with weak demand intelligence can now produce AI-generated content, AI-distributed posts, and AI-personalised emails at ten times the volume — pointed at the wrong audience, in the wrong channel, with the wrong message. At ten times the speed.

The Demand-First Framework

• Map real attention. Before creating anything, understand where your audience actually spends time. SparkToro’s audience research tools reveal the publications, podcasts, and social accounts that your specific buyers actually consume.
• Identify buyer language, not marketer language. The words your buyers use to describe their problems are almost never the words your product team uses to describe their solutions. Ground your content in the actual language of your audience before generating anything at scale.
• Verify category momentum before investing. Producing excellent content in a declining category is a losing investment regardless of quality. Confirm that real buying momentum exists before building.
• Find the trust signals your audience relies on. Identifying which voices, publications, and communities carry authority with your specific audience is the demand intelligence that most AI tools cannot provide — and most teams never gather.

CHAPTER 03

Visibility Is the New Traffic

For two decades, SEO was fundamentally about earning clicks. Rank high, earn a click, bring someone to your site. AI answer engines change that model entirely. When someone asks ChatGPT or Perplexity a question, they receive a synthesised answer — and may never click through to any source at all. Visibility and traffic have been decoupled.

From SEO to GEO: The New Rules of Discoverability

• Original data and research. AI engines are trained to prioritise sources that contain information not available elsewhere. Original surveys, proprietary analyses, and first-party research are the highest-value GEO assets a brand can produce.
• Citability structure. Content must be written so AI systems can extract specific claims, statistics, and answers. Clear headers, short paragraphs, specific assertions, and attributed data all improve citability.
• GEO monitoring. Run your brand name and five core topics through ChatGPT, Perplexity, and Google AI Mode monthly. The gap between what appears and what should appear is your content brief.

CHAPTER 04

Build a Content Engine, Not a Prompt Habit

Ross Simmonds has built his consultancy around one core idea: the best content marketing is not about producing more content. It is about producing content worth distributing. His phrase “create once, distribute forever” captures the system that AI makes newly possible at scale.

The failure mode he sees repeatedly is the “prompt habit”: marketers who use AI to generate individual pieces of content on demand, with no underlying editorial system, no brand voice consistency, and no distribution strategy. The output is fast. The output is plausible. The output is forgettable.

The Content Engine: Three Layers

• The idea layer. Strong content begins with a non-obvious insight that could only come from this brand. AI cannot generate this. It can help develop it once a human has identified it.
• The production layer. Once the core idea exists, AI handles the mechanical work: researching data, drafting the long-form piece, extracting five LinkedIn post angles, writing the newsletter section, scripting the short-form video. One idea becomes seven assets.
• The distribution layer. Content that is not distributed is invisible. Distribution is not an afterthought. It is what makes the production investment worthwhile.

CHAPTER 05

Win Attention Where People Actually Are

Gary Vaynerchuk’s core insight — repeated across a decade of content — is that attention has always been the precondition for everything else in marketing, and that most brands are perpetually late to the channels where attention actually lives.

In 2026: TikTok still offers the largest organic reach opportunity for new entrants. LinkedIn personal profiles offer the highest-quality organic reach for B2B operators. YouTube offers the longest compounding return on investment. Cross-posting content built for one platform into all of them is not a distribution strategy. It is the fastest way to train every algorithm to suppress your content.

Platform-Native Rules

• TikTok (3.70% engagement, +200% brand follower growth): Entertainment first. Hook in 2 seconds. 52% video completion is the benchmark. TikTok Search rivals Google for under-30 product research.
• LinkedIn personal profiles (20–30% organic reach): The last major platform where a human with genuine expertise reaches a large percentage of their network without paid amplification. No external links in post bodies. Company pages reach only 2% of feeds.
• YouTube (the compounding channel): Content created today still drives traffic in five years. Treat it as a search engine. Keyword-first titles. Split-test thumbnails before anything else.
• Instagram (0.48% engagement, -24% YoY): 60–70% Reels for discovery. 20–30% Carousels for saves. No TikTok watermarks.

CHAPTER 06

Turn AI Into a Workflow Advantage

Kieran Flanagan’s argument is that the teams winning with AI are not the ones with the best individual tools. They are the ones who have redesigned their workflows around AI from first principles — identifying every point where a human was doing a task that AI could do as well or better, and systematically removing that friction.

The most common mistake: “spot automation” — using AI to replace individual tasks in an otherwise unchanged workflow. The result is a system that is faster in isolated moments but still fundamentally broken. The teams that win redesign the entire workflow, not just the individual steps.

Workflow Redesign Areas

• Content production pipeline: Research → brief → draft → GEO-optimise → repurpose → schedule → publish. Each step AI-assisted. The human role is editorial judgment at the brief and review stages, not execution.
• Lead qualification and routing: AI scores inbound leads against ICP criteria, enriches CRM records with intent data, and routes leads to the appropriate sales motion before any human touches the record.
• Campaign briefing and variant generation: AI generates the brief, writes copy variants, produces creative options, and recommends the test structure. The execution cycle shortens from weeks to days.

CHAPTER 07

Connect Marketing to Revenue

AI does not automatically close the gap between marketing activity and revenue outcome. In many cases, AI-powered marketing creates a new version of the same problem: faster content production and wider distribution, but no improvement in the quality of leads that actually convert. The work of connecting marketing to revenue is a systems problem, not a content problem.

The Revenue Connection Framework

• Lead scoring with intent data. AI combines behavioural signals with firmographic data to score leads against ICP criteria in real time. This replaces manual qualification and eliminates the “warm body” problem: leads sent to sales before they are ready to buy.
• CRM enrichment and handoff quality. AI enriches CRM records with third-party intent data, competitive research, and engagement history — giving sales reps context they could not have gathered manually.
• Conversion architecture. The conversion path should be designed as a system, not assembled from individual campaigns. AI personalises that path based on the visitor’s industry, role, behaviour, and stage.
• Revenue-linked measurement. Every marketing KPI should be traceable to a revenue outcome. AI-powered attribution is making this more achievable, but the discipline remains a human responsibility.

CHAPTER 08

Onboarding Is Part of Marketing

The promise made in an ad, an article, or a sales call must be fulfilled in the first product experience. If it is not, the acquisition cost was wasted. For digital products, the onboarding experience is the moment of truth.

AI-Powered Onboarding Principles

• Personalise the path to first value. AI segments users at signup based on role, industry, intent signals, and stated goals — and serves a personalised activation sequence for each.
• Reduce time to first value. The single most important metric in onboarding is time to first value. AI removes friction by pre-filling information, suggesting next steps, and surfacing contextual help at the right moment.
• Use email as an onboarding channel. The welcome sequence — a minimum of five emails triggered by signup and activation milestones — should be AI-personalised based on what the user has and has not done.

CHAPTER 09

Retention Is the Real Test

Retention is where the promises made in every earlier stage of the funnel are tested. The true proof of modern marketing is not how fast you acquire customers, but how well you keep and grow them. AI changes the economics of retention in two ways: faster identification of at-risk customers, and more personalised retention interventions.

The Retention Framework

• Churn signal detection. AI models can identify at-risk customers weeks before they cancel, based on changes in login frequency, feature usage, support ticket patterns, and engagement. Early detection gives the retention team a window to intervene before the decision is made.
• Lifecycle messaging. The lifecycle email sequence — triggered by usage milestones, inactivity thresholds, and renewal dates — is the primary retention communication channel. AI personalises this based on each customer’s actual usage patterns.
• Expansion revenue. AI identifies customers showing usage patterns consistent with readiness for a higher tier or additional seat and triggers the appropriate outreach before the customer has actively considered upgrading.

CHAPTER 10

Measure Signal, Not Activity

Christopher Penn’s central argument: most marketing analytics is measuring the wrong things — tracking activity that is easy to count rather than signals that actually predict revenue. The arrival of AI has made this problem worse in a specific way: AI can now generate more reports, more dashboards, and more data visualisations than any team can possibly act on.

The Signal Framework

• Define the outcome first. Before choosing what to measure, define what success looks like in revenue terms. Metrics that cannot be connected to revenue are vanity metrics, regardless of how impressive they look in a dashboard.
• Identify leading indicators. Lagging indicators tell you what happened. Leading indicators tell you what is about to happen. AI is most useful for identifying which leading indicators actually predict lagging outcomes — a correlation analysis most teams have never run.
• Design experiments, not campaigns. Every campaign is a hypothesis, every outcome is data, and every iteration improves the model. AI accelerates the experimentation cycle but cannot replace the discipline of defining the hypothesis before running the test.

CHAPTER 11

Tool Stacks by Stage

The right tool is always the simplest tool that does the required job at the current stage. A solo creator building their first email list does not need HubSpot Breeze. A growth-stage SaaS company managing 50,000 contacts does not need to be on ConvertKit. The tools below are sequenced by stage of growth and matched to documented operator results, not press releases.

BEGINNER Under $100/mo	INTERMEDIATE$300–600/mo	ADVANCED $1,500+/mo
Claude or ChatGPT — content, copyPerplexity — researchCanva AI — visualsBeehiiv / ConvertKit — emailBuffer or Later — scheduling Master prompting before adding tools.	Claude — copy and ideationSurfer SEO or Frase — SEO/GEOMidjourney — visual creativeActiveCampaign — email automationOpus Clip — video repurposingn8n or Make — workflow automation What a 3-person team did 5 years ago.	Claude + Jasper — content engineHubSpot Breeze — CRM + agentsGoodie AI — GEO monitoringSeventh Sense — email timingRunway + Descript — videon8n — agentic pipelines Full agentic marketing stack.

CHAPTER 12

The New Marketing Moat

Every industry transformation produces two kinds of operators: those who see the structural shift clearly enough to reorganise around it, and those who add the new technology to the old model and wonder why the results are underwhelming.

The organisations building durable marketing advantages in 2026 are not the ones with the most sophisticated tool stacks. They are the ones who understood, early enough to act on it, that AI changes the cost of execution — not the value of genuine expertise, not the power of authentic audience relationships, and not the irreplaceable quality of a human perspective that makes someone trust a voice enough to follow it.

The Four Moats That Survive the AI Era

• Audience relationships. A list of 30,000 email subscribers with 40% open rates cannot be reproduced by a competitor with a better AI stack. No tool can generate it. It can only be earned.
• Original data and research. Any AI can synthesise existing public information. No AI can produce data that does not exist yet. Operators who generate original research have a GEO moat that no competitor can buy.
• Genuine domain expertise. The operator who has genuinely done the thing has a signal that AI cannot fake and that the most sophisticated algorithms are specifically designed to detect.
• Speed of learning. The organisations that compound fastest are not those with the most tools but those with the most sophisticated feedback loops. Strategic intelligence remains exclusively human.

“The stack is the infrastructure. The moat is what you build with the time the stack gives back to you.”