The Factory Nobody Talks About

Everyone wants the chocolate river.

They want the personalized email that references what a prospect said on a sales call two weeks ago. They want the ABM landing page that reads like it was written for one company and one company only. They want the case study that turns a 30-minute customer interview into a blog post, a LinkedIn carousel, a set of sales talking points, and a quote library. Much like the late and great Freddie Mercury sang decades ago, they want it all, and they want it now.

The magic of what AI makes possible is what sells. But nobody wants to talk about the pipes.

• The workflows that connect a CRM record to a content engine.
• The structured data layer that lets a sales call transcript auto-generate a follow-up email with the right value prop attached.
• The tagging system that means a new rep can search for “enterprise objection handling” and get thirty real customer quotes in two seconds.

Even with AI, without the right pipes in the right places, the chocolate river simply doesn’t flow.

Most people think of automation as the Henry Ford model. A grim assembly line cranking out identical Model Ts. Each unit is the same as the last: rigid, hyper-efficient, and painfully predictable.

We sometimes forget, though, that not all factories are created equal.

Think about it: if Henry Ford built a factory for repetition, Willy Wonka built a factory for imagination. Wonka’s factory didn’t produce just one thing. It produced Everlasting Gobstoppers and Fizzy Lifting Drinks and chocolate rivers. Every room was different, and every output was unique.

But all of it moved through pipes.

The chocolate rivers of the world get all the attention. The creative outputs, the personalized content, or the marketing campaigns that actually feel human. It’s easy to overlook the fact that none of it exists without the infrastructure underneath.

Ford automated for sameness. Wonka automated for creativity.

Most teams in 2026 aren’t building either kind of factory. They’re doing everything with AI chat tools, one prompt at a time.

§The State of Go-to-Market in 2026

The picture isn’t complicated, but it is uncomfortable.

Marketing teams are smaller than they were three years ago. Budgets fell from 9% of revenue to 6% in two years, while output expectations kept climbing.¹ The teams that survived layoffs inherited the work of the people who didn’t.

The technology stack keeps growing in the opposite direction, too. The average B2B organization runs 12 to 20 marketing tools.² The martech market nearly doubled in five years, from roughly 8,000 tools to more than 15,400.³ Gartner reports that only 49% are actively used.⁴

Half the tools, sitting there, burning budget, doing nothing (all, ironically, in the name of efficiency).

But here’s the number that matters: 51% of B2B organizations that implement AI fail to achieve the outcomes they expected.⁵ Meanwhile, 88% of organizations now use AI in at least one function,⁶ and only 35% of marketing teams have the infrastructure to make it work.⁷

Nearly everyone has adopted AI. More than half aren’t getting what they expected. About a third have the infrastructure to actually make it work.

That gap is what this book is about.

§The Difference Between Using AI and Building With AI

There are two ways companies relate to AI, and the gap between them explains most of the frustration.

The first is using AI. You open ChatGPT, write a prompt, get an output. Summarize a sales call. Draft a blog post. Rewrite an email. Each task is standalone. You get a 20% efficiency gain on that specific thing, then start from scratch on the next one.

This is where most companies are. Sopro’s 2025 analysis found that 94% of marketers now use AI in some form,⁸ which sounds impressive until you realize that for the majority, “using AI” means asking it to do one thing at a time.

The second is building with AI. You chain tasks into workflows where the output of one step becomes the input for the next. A sales call transcript doesn’t just get summarized. It flows through a system that extracts pain points, maps them to your value propositions, generates a personalized follow-up email, creates a custom one-pager, tags recurring themes for the content team, and stores the insights in a searchable database.

One input. Multiple outputs. Connected to every other department in your organization.

Using AI shaves time off individual tasks. Building with AI restructures the economics of your entire operation.

§The Buyer Isn’t in Your Funnel

For years, B2B marketing was built on a linear model. It assumed that someone found your content, filled out a form, entered the funnel, got nurtured by email, talked to sales, and bought your product. Every tool and every metric and, eventually, every org chart was designed around this progression.

The typical buying group for a complex B2B solution now involves six to ten decision makers.⁹ Forrester puts the average at 13 stakeholders, with 89% of buying decisions crossing multiple departments.¹⁰ Roughly 80% of the B2B buying journey happens without direct vendor contact.¹¹ Buyers spend about 17% of their total buying time meeting with potential suppliers, and that time is split across every vendor they’re evaluating.

This means that the average buyer’s journey is more akin to moving through a maze than through a funnel. They see your blog post on Monday, a competitor’s G2 reviews on Tuesday, a question in a Slack community on Wednesday, an AI-generated summary of your product category on Thursday, and hold an internal meeting about budget on Friday. They might loop back to your website three times before ever filling out a form, and 75% say they’d prefer a rep-free experience entirely.¹²

You can’t control the sequence anymore. But you can build a system that produces the right content, in the right format, across every touchpoint the buyer might hit. Plus, you can now do all of that without a 15-person team, if the infrastructure is right.

§Strengths and Weaknesses of Content-Led and Product-Led Growth

§For the last decade, B2B companies organized growth around a few ideas, but two really took flight.

1. Content-led growth: create great content, drive traffic, capture leads, nurture them. It worked when content was scarce, Google was the primary discovery channel, and a good blog post could rank for years. At scale, a content-led team could be 15 to 30 people.

That model is breaking, if it’s not already broken. Content is no longer scarce. AI can produce a passable blog post in seconds, and every company on earth now publishes. Gartner predicts traditional search volume will decline 25% this year as AI-powered tools capture share.¹³ eMarketer data indicates 80% of B2B buyers now use ChatGPT and Perplexity as much as Google when researching vendors.¹⁴

The signal-to-noise ratio has also collapsed (i.e. more content with fewer readers). The readers who do show up increasingly got there through an AI intermediary that may never have sent them to your website at all.

2. Product-led growth: let the product sell itself through free trials and self-serve onboarding. It worked when software was differentiated and building a good product was hard.

But with advancement in AI coding, that model is under pressure too. People can scoff at the term “vibe coding” all they’d like, but it doesn’t change the fact that AI makes it easier to replicate features. When a small team can build in weeks what used to take months, “the product sells itself” stops being a strategy and becomes a race to the bottom.

Both models share the same blind spot: they treat growth as a function of one thing. But modern buyers don’t move through a content funnel or a product funnel. They move through a maze of touchpoints, conversations, AI-generated answers, and internal politics.

§What I Learned Running a Lean Growth Engine

Before diving too far into the deep end, I should tell you how I got here, because this book is more documentation than theory.

In early 2022, after years of doing SEO for bootstrapped startups, I was working at a content agency, writing long-form posts and landing pages for SaaS companies. One of my clients was Copy.ai. I wrote their first four landing pages, which performed well enough that the company noticed. By the end of that year I’d gone from agency contractor to freelancer to full-time employee.

Copy.ai was one of the first companies to ship Workflows in 2023. At that time, most marketers were still resistant to the idea that AI could take their job. They didn’t want to flirt with Chat tools, let alone build out any meaningful processes with AI. But internally, at Copy.ai, workflows introduced multi-threaded prompting: chaining tasks together so the output of one became the input of the next. Instead of asking AI to do one thing, I could actually build out an entire process where each step informed the next.

That distinction changed everything.

When the AI boom exploded in 2023, Copy.ai’s free tools drove hundreds of thousands of monthly visits, which sounds like an SEO’s dream. The problem was that none of it was converting into enterprise contracts. The traffic was students, hobbyists, and people Googling terms like “sentence rewriter” or “instagram caption generator” from around the world. The stats looked great on a dashboard, but they meant nothing to pipeline.

In 2024, with the arrival of a new (and very talented) CMO, my mandate became turning that B2C traffic engine into a B2B pipeline machine.

I deliberately killed pages driving tens of thousands of visits because they attracted the wrong people. I cleaned up technical debt and rebuilt the content strategy around ICP-focused pages.

Traffic went down from 350k to 210k monthly visits over the course of 18 months. But our enterprise pipeline went from effectively zero to multi-millions, with ARR reaching $3-4M (again, from enterprise contracts, not self-serve students in other countries).

Most people would look at those traffic numbers and think something went wrong. This even came up on a few pre-sales calls from marketers who wanted to show their boss, who had set the meeting, that AI couldn’t produce high-quality content. But that decision, choosing pipeline over pageviews, is at the core of everything I believe about growth.

In February 2024, I gave a talk at a content marketing conference, about using AI in content strategy. The room was hostile, or at least it felt that way at the time. Experienced content marketers who had built their careers on the craft of writing, and I was telling them the way they worked was about to be restructured.

Needless to say, I didn’t make a ton of friends at that conference.

In most cases when very intelligent people disagree with me (and the room was full of very intelligent people), I default to the idea that I must be missing something. It can’t be them, so what am I missing? So I went home, put my beagle on the leash, put my headphones in, and set off to figure out what it was that they knew that I didn’t.

But after a few days and many miles of walking Zoomer (our beagle), for the first time in my career, I was certain: I was right and they were wrong. Not about AI replacing writers, which was never my argument. I knew that I was right about systems, not individual talent, determining who wins, and that AI would have a major impact on how operational systems shaped SaaS.

That conviction hasn’t wavered. If anything, it’s only gotten stronger.

§What Systems-Led Growth Actually Means

Pipeline over pageviews. A skeleton crew doing the work of a department. Workflows that turn one input into ten outputs. All of it pointed me toward a single idea: the companies that do best over the next decade won’t be the ones with the biggest teams or the biggest budgets. They’ll be the ones with the best systems.

I call that Systems-Led Growth.

A sales call gets recorded and transcribed. The transcript flows through a workflow that extracts pain points, maps them to your value propositions, and produces a personalized follow-up email, a custom one-pager, and talking points for the next call. The recurring themes get tagged and stored, so when marketing needs to write a blog post, they’re pulling from the actual words buyers are using.

A podcast episode gets recorded. The transcript flows through a workflow that generates a thought leadership article, a LinkedIn post, a newsletter draft, a YouTube description, a landing page, and social clips. One conversation becomes ten assets and nobody at the organization starts from a blank page.

None of this requires a large team, but all of it requires a system built by expert practitioners.

• A prompt is a task: you ask AI to do one thing.
• A workflow is a process: you chain tasks so they build on each other.
• A system is a growth engine: connected workflows where every conversation, every interaction, every piece of customer insight turns into compounding assets across your entire go-to-market.

Systems-Led Growth sits alongside content-led growth and product-led growth. Not as a replacement for either, but as the connective layer that makes both work. Content is still important. Your product still needs to be good. Neither is sufficient on its own.

§The Four Principles of Systems-Led Growth

1. Systems over prompts

A single prompt is a task. A chain of connected workflows is a growth engine. The best prompts won’t win this era. The best systems will.

2. Pipeline over pageviews

Traffic that doesn’t convert is vanity. The hardest and most valuable decision in growth is killing something that looks good on a dashboard but doesn’t drive business outcomes.

3. Skeleton crews can win

One person with the right architecture can outperform a 15-person marketing department.

4. Lived experience is the only moat left

AI can generate any content, but it can’t say something that only you can say. In a world of infinite content, the defensible position is specificity rooted in experience.

§What This Book Is and What This Book Isn’t

This is the second edition of Pipes Before Chocolate. The first was a short book I wrote while building growth systems at Copy.ai: the Wonka metaphor, the workflow-first approach, and the argument that infrastructure matters more than creative output.

This edition is bigger, more tactical, and pressure-tested by another year and a half of building.

Part One establishes the framework: the problem, the metaphor, the three levels of AI usage, and the role humans play in the loop. Part Two is the tactical core, one chapter per go-to-market function: organic content, sales outbound, inbound processing, thought leadership, ABM, events, case studies. Every chapter follows the same structure: the problem, the old way, the system, the workflow, what to measure, and one honest admission about what doesn’t work yet. Part Three is what happens when the pipes are in place: how the system compounds, how to measure it, and where, I believe, this is all going.

What this book isn’t: a beginner’s guide to AI (I’m going to assume you’ve used ChatGPT or Claude at least a handful of times), a tool comparison (tools change fast; opinions about them have a short shelf life, and they’re getting shorter), or a book about prompting. This is about what happens after the prompt, when you need to connect the output to everything else your business is trying to do.

Every system I describe, I’ve built. Every number I cite from my own experience is real. Where I’m still figuring something out, I’ll say so. Where something didn’t work, I’ll tell you that, too.

And no, I’m not a self-proclaimed AI guru. I’m a practitioner who documents what he builds. The systems in this book are the ones I use every day: managing SEO, building content engines, developing AEO frameworks, trying to grow businesses while raising two kids in Cochrane, Alberta, and keeping my dog both well fed and well walked.

Time is my scarcest resource. If you’re reading this, it’s probably yours too.

§Laying the Pipes First

The chocolate will come. The personalized campaigns, the one-to-one customer experiences, the ABM sequences that feel hand-written. All of that is possible now, with the tools we have, for teams of any size.

But it starts with the pipes.

Most companies are still debating which AI tool is best. The ones that genuinely pull ahead this year will be the ones building the system while the debate continues.

That window is still open. It won’t be forever.