• LinkedIn
• Twitter
• Instagram
• YouTube
• RSS Feed

Apple •Spotify• Pocket Casts •Youtube •Overcast •RSS

What’s up folks, welcome to our 4 part series of Crawling through the dungeon of martech architecture. You’ve arrived at Part 1 : The Fall of CRM Gravity.

Summary: At some point in the last decade, the CRM became a shared apartment with 19 roommates, each adding their own version of the source of truth. This episode argues for a cleaner path: raw data into the warehouse, transformation with tools like dbt, then activation through reverse ETL so definitions stay centralized and audiences stay portable. The real villain is the Export Hydra: copying warehouse data into every tool until every copy drifts. Zero-copy sharing breaks that reflex, and AI makes the stakes obvious. Every agent reads from somewhere. If it reads from copied junk, it inherits the mess.

In this Episode…

• Why the CRM Lost Its Authority
• BOSS BATTLE: The False Truth King
• Why Centralizing Data Only to Copy It Out Defeats the Purpose
• BOSS BATTLE: The Export Hydra
• How to Move to a Warehouse-Native Architecture
• How to Achieve Portable Audiences
• How CLI/MCP Servers Are Changing Marketing Stack Integration

Welcome to the descent into the Dungeon of Martech Architecture, a 4-part journey through the unhinged and constantly expanding universe of marketing technology.

As a massive sci-fi fan currently reading the Dungeon Crawler Carl books, I have used their level-by-level progression as the direct inspiration for this ‘dungeon crawl’ analogy, and while you don’t need to know the books to enjoy the journey, those who do will recognize some of the gaming lore and achievement-style rewards woven into our descent. 

This will be educational and helpful for anyone that works and builds martech, and hopefully it’s also a bit fun. Without a doubt though, it will be weird. 

Here is your quick guide to the floors ahead:

Episode 1: CRM Gravity. You’ll conquer the source of truth and discover that the data warehouse replaces the CRM with portable audiences.

Episode 2: The Eye of Context. You’ll learn why AI fails without shared meaning, why context engineering is the layer between data and agent authority, and why the industry built the wrong kind of meaning infrastructure in 2012.

Episode 3: The Correlation Masquerade. You’ll escape the correlation trap and build the causal memory layer that separates agents that optimize correctly from agents that confidently scale the wrong behavior.

Episode 4: The Dispatch Tower. You’ll tackle the governance chaos of 30 vendors all claiming authority, and confront the interface decision that most organizations already made without realizing it.

Let’s start our descent.

Be honest: when was the last time you pulled up a number in your CRM and actually trusted it? like… no second-guessing, no “that feels a bit off”… just total confidence?

Maybe you didn’t really have time to double check the logic behind the number and you were too excited to share the positive results. So you forwarded it to a peer. 

Or maybe you’ve been in that meeting… 2 people arguing over a number, both pull it up in the same CRM, and somehow get 2 completely different answers… and no one can explain which one’s actually right.

We’ve all been there, we’ve felt it. That dark, creeping dread. When “which number is right?” gets answered with “well… it depends who built the report,”. They know it. You know it. The CRM admin knows it. Everyone in the room knows it. You don’t have a source of truth… just a CRM that’s turned into a dumping ground of lost updates that have slowly compounded into competing versions of reality.

Call it counterfeit truth or data mirage… I call it bad data. Data that has the appearance of authority without the actual authority behind it. It’s everywhere in the modern marketing stack. And the CRM is often where it starts.

That’s where our first boss is hiding. 

FLOOR 1: Why the CRM Lost Its Authority

If you’re in B2B or B2C the first floor looks a bit different but only because of terminology. In B2B the 2 cornerstone platforms are the CRM and the MAP: the Customer Relationship Management software and the Marketing Automation Platform. Sales works in the former, marketing works in the latter, ops is stuck making the two talk to each other.

In B2C though, for some reason you all decided that the MAP is actually called a CRM and the B2B version of the CRM isn’t really needed because there’s often no sales team, instead it’s a customer support or product led motion.

	B2B	B2C
Primary platforms	CRM + MAP Separate tools	CRM Functions as MAP
CRM owner	Sales	Marketing / support
Motion	Sales-led	Product-led or support-led
Ops job	Making CRM and MAP talk	Managing the single platform

In both scenarios though the same thing happens to that central platform. It gets inherited by teams that weren’t its original audience. It accumulates data it wasn’t designed to hold. And it becomes the unofficial source of truth for the whole business without anyone explicitly deciding that was a good idea.

Why Every Team Moved Into the CRM (And How It Lost Its Authority)

So how did we get here? CRMs were built for one job: tracking the sales motion. Contacts, deals, stages, activity logs. They were good at that job. Then marketing moved in. Marketers ruin everything. But leadership is worse. Leadership started pulling board metrics from the CRM. Then the product team added usage data. Then we added ABM and account signals, and we had to push that data somewhere. Then AI interactions needed a home.

What a freaking mess.

Everyone needed a record of the customer, and the CRM was already there. It’s literally called the Customer Relationship Manager. So it became the shared folder everyone saved their customer work into, even though it was designed for a very specific kind of work.

The problem is that once data is stored in a CRM, it starts reflecting the team that works there. Sales edits the contact. Marketing overwrites a field. Customer success adds a note. Each edit is local logic applied to what everyone assumes is shared truth. The data looks official but you know deep down that the authority behind it belongs to whoever edited it last.

Meg Gowell, Head of Marketing at Elly.ai and former Head of Marketing at Typeform  crossed over from a Salesforce-first organization to one where the warehouse had already taken over:

“Coming into Typeform, I was used to Salesforce or HubSpot being the source of truth. But here, the core business is represented more in the data warehouse, and Salesforce supports the sales-led side. That makes the data foundation really important, because in PLG, self-serve and sales-led have to serve each other.”

MEG GOWELL, Episode 155

At Typeform, the CRM just stopped being the brain. The warehouse had already taken that role, and once you’ve seen that transition, it’s very hard to unsee it.

Istvan Meszaros, Founder and CEO of Mitzu.io, who built warehouse-native analytics at scale, argues that the CRM was always the wrong anchor:

“CRM is often treated like the source of truth, but it can feel more like a glorified spreadsheet. The warehouse gives you a more visible, reviewable process. Mistakes can still happen, but at least you can trace how they happened, fix the job, and stop pretending every manual field update is trustworthy.”

ISTVAN MESZAROS, Episode 180

Access and version control has long been an issue with CRMs when you start letting other teams move in. Change a lead status, bulk upload a list, manually tweak a lifecycle stage, and nobody questions it. Nobody remembers who made the change or why. Warehouses demand structure: updates happen through scheduled jobs, written in code, version-controlled, deployed with intention. The friction is the feature in this case. And the whole thing is kept under lock and key. 

Here’s David Joosten, Co-Founder at GrowthLoop and the co-author of ‘First-Party Data Activation’. He’s seen this from both sides, platform builder and enterprise architect, and he identifies the 2 pain points marketers actually care about:

“Marketers usually don’t care how the sausage gets made until something breaks. They care when the wrong audience gets a campaign, and they care when everything feels too slow. Reliability matters because it solves both problems.”

DAVID JOOSTEN, Episode 193

Marketers may not care how the sausage gets made, but they care deeply when a campaign hits the wrong people, uses stale data, or moves too slowly to matter. David expands on this and he picks on Marketo a bit:

“Systems like Marketo can be really opaque from a data perspective. You can send an update and look up one record, but you don’t always get the observability to know every event made it in cleanly. That’s the gap the warehouse solves much better.”

DAVID JOOSTEN, Episode 193

Why the CRM Became a Dump of Counterfeit Truth

CRM Data Loses Integrity When It Travels

When the answer to “can we share this with finance, product, or leadership” is always “let me pull a report and send a spreadsheet,” the system is the source of truth for one team only, with a UI that’s often really ugly. 

Here’s David Joosten explaining why the tool itself is part of the problem:

“Marketing automation tools are usually optimized for a specific motion. Marketo thinks in B2B. Braze and Klaviyo think more in B2C. But when your business has agencies, SMBs, enterprise customers, and multiple product lines, it’s unlikely one marketing tool can reflect the full truth of the business.”

DAVID JOOSTEN, Episode 193

Both David and Istvan are fantastic pioneers in this space, and yes, their perspective comes with a bias. But it is the earned kind of bias. They did not wake up one day, build a tool, and then reverse-engineer a philosophy around it. They built a company around this belief because they were already convinced the warehouse should play a more central role in modern marketing systems.

Let’s hear from someone who saw this from the other side of the stack.

Kevin White, now Head of Marketing at Scrunch.ai, spent years at Segment during its early rise. He was close to the infrastructure story as it unfolded, and now has enough distance from it to reflect on what actually happens when data starts moving downstream:

“I think the warehouse becomes the source of truth because it’s the only place that can handle the full complexity. Once data hits downstream tools, it starts to drift. Salesforce becomes one endpoint among many, while marketers shape and enrich the data in the warehouse and push it wherever it needs to go.”

KEVIN WHITE, Episode 183

I love that Kevin uses entropy in his argument. Data that travels to downstream tools loses integrity over time and the CRM compounds those mistakes rather than containing them.

The warehouse solves the entropy problem in a different way. Every transformation is written, reviewed, and deployed explicitly. There’s no “someone must have changed this at some point.” Every change has an author, a timestamp, and a reason. There’s auditability and version control. That is a different relationship between a team and its data; one where the system resists drift instead of accumulating it.

	CRM	Data Warehouse
How data changes	Manual edits, bulk uploads, UI fields	Scheduled jobs, written in code, version-controlled
When something’s wrong	Hard to detect, no audit trail	Easier to catch; jobs can be reviewed
Who changed it?	Not always obvious. Whoever ran the last import	Always has an author, timestamp, and reason
Trust model	Appears authoritative	Earns authority through structure

Why CRM Gravity Outlasts the Technical Argument

Obviously we can’t go deep into the dungeon of martech without hearing from the legendary figure who built the tunnel by hand. But before we hear from Scott Brinker, I did want to touch on a recent report he worked on with Databricks. 

Bryce Peake, VP of Digital Products and Technology at Domino’s, is quoted in the report:

“We’ve been trying to modify the same martech stack we’ve had since the internet started interneting. Folks, we’re going to have to build a new one.”

Bryce Peake, VP of Digital Products and Technology at Domino’s

That’s a Fortune 500 brand with decades of infrastructure saying the patch has run out of runway.

The CRM is so hard to dethrone because teams build their identity around it. I’ve worked at companies that were so indoctrinated into SF that their mandate from the top was “If it doesn’t live in SF, it doesn’t exist”. This was like a decade ago and since then it’s impossible to argue there are better tools for different types of information but I still see this in companies today and the idea of crossing to a warehouse first model is just super foreign.

Meg Gowell, describing her own experience crossing into a warehouse-first organization for the first time, captures what’s really happening:

“I never thought I’d miss Salesforce reporting, but moving into a warehouse-first world has been a real learning curve. There’s learning Looker, and then there’s learning the fields, definitions, and data structure underneath it. The upside is there’s more data available than I ever had in a CRM. The hard part is rebuilding the familiarity and learning how to spot-check whether the data is actually telling the truth.”

MEG GOWELL, Episode 155

The CRM feels like the center because that’s where the people doing the work are. The sales rep works in Salesforce. The marketing manager works in HubSpot. The lifecycle Manager in MoEngage. Data engineers work in the warehouse. Getting the marketing team to trust the warehouse means getting them to trust a system where their day-to-day work doesn’t happen. That resistance is human. And it is the real reason CRM gravity persists long past the point where it should have given way.

John Saunders, VP of Product at Power Digital Marketing, spent years building nova, an internal operating system designed to give the agency a single source of truth. He learned the hard way that the technology was the easier half of the problem:

“The source of truth only works if the organization actually believes in it. It’s less of a product problem and more of a people problem. If teams aren’t aligned on where truth lives, you don’t have a source of truth. You have another tool.”

JOHN SAUNDERS, Episode 187

Nova’s early years focused on consolidating everything into one platform: Salesforce records, client service metrics, ad spend data. In practice, it created a different problem: people didn’t want the data, they wanted a quick story of what changed and why. The organizational agreement had to come before the data engineering could matter. Without it, the platform was just another system people worked around.

Okay let’s recap what this floor contained. 3 separate failure modes, all rooted in the same thing.

1. The CRM was built for the sales motion and never rebuilt for everyone who moved in after.
2. Every field looks authoritative, but the authority belongs to whoever ran the last import.
3. The teams most anchored to the CRM are the ones most resistant to trusting the warehouse, because the warehouse is not where their day starts.

Counterfeit truth gets its grip from the humans who never agreed to stop trusting it.

BOSS BATTLE: The False Truth King

The boss on this floor that you need to defeat is The False Truth King: data with the appearance of authority and none of the substance behind it. 

The key to defeating this boss is understanding what the CRM was always actually good at (managing the sales motion) and what it was never built to be: the center of marketing truth, let alone organizational truth. So our end goal isn’t to kill the CRM, I’m still a big fan, but we want to demote it in a way. The end goal is letting the warehouse take the brain.

*Inventory check.*

Before you face the boss, here’s what you picked up on this floor.

[LEGENDARY] The Data Warehouse: Every change has an author, a job, and a timestamp. Updates happen through scheduled, version-controlled code; every change is traceable. Counterfeit truth cannot survive a system that audits everything by default.

That’s the loadout for this floor. The boss is ahead, should be easy to defeat.

Right?

Not so fast, martech crawler. Knowing how to defeat this boss is one thing, but actually getting that thing implemented is actually the bigger battle. Because human decision makers are involved and a defiant marketer proclaiming the DWH needs to exist needs way more than just a proclamation. 

The transition from CRM to DWH is harder to sell than it sounds. Istvan Meszaros, who has watched teams make this shift, explains why it’s an uphill battle:

“For the end user, this shouldn’t feel like a new category. They should still be able to do the same analytics they’re used to. The real shift is for the team building the marketing infrastructure. For them, it’s a completely new approach.”

ISTVAN MESZAROS, Episode 180

Lots of politics at play here. The revolution happens at the infrastructure layer, in a language that doesn’t translate upward. The CMO doesn’t experience the change. The marketer doesn’t see the architecture shift. The person who has to fight for it has to win a budget conversation about a thing the beneficiary will never notice.

But it’s doable… I believe in you!

The approach that works is the same as any political change: find the team that’s already in pain. The analyst who can’t answer the same question twice and get the same number. The ops person manually exporting CSVs every Tuesday. Win there first, then let that win do the selling.

“It’s much more effective to have a few passionate champions than a crowd of lukewarm supporters. You almost have to treat it like a small political campaign: find the people who care deeply, then let them help you win over everyone else.”

DAVID JOOSTEN, Episode 193

You might not be able to win this boss fight with an abstract infrastructure argument during your RevOps budget meeting. But you can defeat it with a before-and-after story from a team that stopped fighting about numbers.

This might actually be a months-long battle depending on the resources and the team’s priorities. But it’s doable, especially in the age of AI. I’ve had success getting this rolled out as a dependency to all the AI projects the execs want to launch. 

But the bad news is that this boss doesn’t end with just shining a light on the counterfeit truth. Counterfeit truth at the source is only the first form of the boss. Here’s the second.

Let’s say you heroically convinced the organization to build the warehouse. The exec team is pumped about a single source of truth. You go through an extensive implementation journey and on the other side, the data is clean, governed, and centralized. Then someone asks the data team to use it to run a campaign.

Absolutely, no sweat. Here’s your audience. Where do you want this?

The second form of this boss is a vault you spent a year building, where every exit leads to a photocopier.

Why Centralizing Data Only to Copy It Out Defeats the Purpose

Think about what that means. You invested the organizational capital, the engineering hours, the months of stakeholder negotiation to build a central source of truth. And now you’re making copies of it and distributing them to every downstream tool in the stack, where each copy immediately begins drifting away from the original.

It is like building a data treasury and then immediately photocopying everything inside it and mailing the copies to different departments. 

Here’s Lourenco Mello, Director of Product Marketing at Snowflake on the paradox:

“You do all this hard cultural work to unify your data and create a source of truth, then somehow the next move is to copy that data back out for AI modeling or campaign execution. That feels backwards. The better path is the applications coming to the data, with fewer copies and less drift.”

LOURENCO MELLO, Episode 142

This isn’t just a B2B thing either. You could actually make the case that it’s even more prevalent and important for big B2C orgs. What I hear from the ground talking to B2C leaders is some version of:

• I’ve centralized a complete picture of my customer, we’ve got purchase behavior and product engagement to support interactions and marketing touchpoints, all living together in my data warehouse.
• I want to activate that warehouse data directly and turn it into real-time, highly relevant, deeply personalized experiences across my campaigns.
• I can’t afford stale data or the mess and friction that comes from copying it across a dozen different tools.

Same diagnosis, across markets, across org types. The warehouse holds the truth. The tools need to come to the data.

Here’s Erin Foxworthy, Snowflake’s Global Industry GTM Lead, Marketers and Advertisers, describes what this shift looks like from the infrastructure side:

“If you’ve done the work to build your marketing foundation in Snowflake, the next wave of applications should come to that data, not pull it somewhere else. The goal is to collaborate across the ecosystem while moving data as little as possible, because that’s better for cost, security, scale, and trust.”

ERIN FOXWORTHY, Episode 138

That’s the flip from the old model. Instead of data pipelines pushing data outward to wherever it’s needed, platforms are building native connectors that read directly from the warehouse. The mechanism Erin describes is data sharing (what some also call zero-copy) the ability to expose a live view of warehouse data to a downstream tool without ever moving it. No export, no sync job, no copy that immediately starts drifting. The data stays where it is, and the application comes to it.

The enforcement of this model is already happening at the buying level. One major holding company, Erin said, went to market with a firm position: they would only accept data through a data share going forward, and they were using their buying power to make it stick. When procurement starts enforcing the architecture, the technical argument has become a commercial one.

BOSS BATTLE: The Export Hydra

The Export Hydra: the assumption so embedded in how marketing teams operate that it never registers as a choice. Data has to travel to wherever it’s needed. You pull it, clean it, load it into the tool. That’s just how it works. But every export is a copy. Every copy starts drifting. 

Defeating this form means stopping the copies entirely.

For this battle, you have 4 items.

*Inventory check.* 4 items before this form.

[LEGENDARY] The Data Warehouse: Every change has an author, a job, and a timestamp. The foundation that makes every other item work. (Acquired: Form 1)

[RARE] dbt: Transforms data inside the warehouse without moving it. Converts raw ingestion into governed, trustworthy signal.

[RARE] Reverse ETL: Pushes governed data to downstream tools from a single source of truth. 1 definition, everywhere it needs to go.

[EPIC] Zero-Copy Data Share: Downstream tools read from the warehouse directly instead of receiving exports. Nothing gets photocopied. Nothing starts drifting the moment it leaves.

How to Move to a Warehouse-Native Architecture

Teams that have cleared this floor and beat the The Export Hydra boss describe their victory the same way: 3 steps, in the same order every time.

Danny Lambert, the former Director of Marketing Operations at dbt Labs, has built this stack from scratch at multiple companies, gave us the clearest version of the blueprint:

“The roadmap is pretty simple: extract your data into a cloud warehouse, transform it there with something like dbt, then use reverse ETL to push the cleaned-up data where it needs to go. That’s the basic path from scattered tools to a usable marketing data foundation.”

DANNY LAMBERT, Episode 143

That’s the skeleton for teams starting fresh. 

1. Centralize: Funnel all raw data into a cloud data warehouse. Use extraction tools (Stitch, Fivetran) to continuously load from all sources into one place: Snowflake, Databricks, Redshift, BigQuery.
2. Transform: Clean, join, and model that data inside the warehouse. dbt lets you do the transformation where the data already lives, instead of moving it somewhere else to shape it.
3. Activate: Push refined data out to the tools that need it. Reverse ETL platforms like GrowthLoop and Hightouch sync the warehouse into your CRM, marketing automation, and ad platforms, wherever execution needs to happen.

For teams already deep in marketing automation and CRM, here’s David Joosten describing the migration path he’s seen work:

“The bridge is to rebuild the warehouse around the marketing schema people already trust. Don’t rip away their beloved source of truth and tell them the warehouse wins now. Match what they know, prove the data lines up, then start feeding marketing automation from one trusted foundation with better reliability, observability, and alerts when something breaks.”

DAVID JOOSTEN, Episode 193

So that’s the skeleton to beat The Export Hydra boss complete that migration. 

How to Achieve Portable Audiences

Somewhere along this floor though, we encounter a few operating realities that become non-negotiable:

• Define audiences once, not separately in every tool.
• Reuse the same segmentation logic across every channel.
• Minimize the number of handoffs where data can silently drift.

For all the complexities of the long B2B sales journey and account roll-ups and all that crap… The B2C folks deal with a challenge on this floor that most B2B marketers only dream of… and that’s volume.

One audience definition, on every channel, at the scale of 30M user records… is a whole different ball game than your 300k lead database in B2B. 

Here’s Hope Barrett, Senior Director of Product Management for Martech at SoundCloud, on how her team solved this:

“Everything goes into BigQuery because that’s our source of truth. I want audiences built once in the warehouse, then pushed wherever they need to go, instead of relying on every platform’s native integrations. I don’t want to build the same audience five times. That just adds all sorts of room for error.”

HOPE BARRETT, Episode 175

One place, one audience definition, and their reverse ETL sends it everywhere. The native integrations and vendor SDKs get bypassed entirely, because every one is a potential drift point.

Here’s Blair Bendel, Senior VP of Marketing at Foxwoods, inherited the opposite situation, with data fragmented across disconnected systems by industry and business unit and no common layer underneath:

“We have gaming data, hotel data, food and beverage data, and a lot more across the business. The challenge is bringing all of that together in a way the team can actually understand and act on. Because if you have all this data but can’t execute on it, it becomes counterproductive pretty quickly.”

BLAIR BENDEL, Episode 196

The vocabulary is different across those 2 stories. Hope is building audiences in BigQuery and deploying through reverse ETL into MoEngage. Blair is consolidating gaming, hotel, and food-and-beverage data into a single execution layer. 

The infrastructure they’re describing is the same: one definition of the customer, governed in one place, available everywhere the business needs it. And the outcome is the same: the ability to execute across channels without rebuilding the customer profile from scratch every time.

The thread connecting both stories: omnichannel execution is a data architecture strategy. The channels don’t matter if the foundation beneath them is fragmented.

Here’s David Joosten (one last time in this episode) on what that architecture gives the marketer:

“Self-serve targeting changes the whole rhythm for marketers. If they can test filters, see audience size, understand spend, and shape personalization ideas before the data team gets involved, the path from idea to activation gets much shorter. And marketers love that feeling of moving faster.”

DAVID JOOSTEN, Episode 193

It feels like for years, this architecture was justified on reporting and efficiency grounds: one source of truth, fewer sync jobs, cleaner pipelines. Those arguments were always true, but in revenue hungry prioritization sessions, they never won. Data warehouse investments got funded by data teams and engineering.

But in 2026, the conversation is different. You could go as far as saying that, if you want to do this properly, the warehouse-first architecture is now the prerequisite for AI. Every agent that runs on top of your stack reads from something. If what it reads from is a patchwork of copies and downstream exports, the agent inherits all of that drift, all of that latency, all of that contradictory logic. 

If what it reads from is a governed, centralized foundation with consistent definitions and controlled access, the agent can actually do something trustworthy with it.

It’s the comeback of data quality. The argument that never moved marketing budgets finally has a forcing function. The funniest part is that the architecture and the whole modern data stack actually hasn’t evolved that much from a foundation stand point. What’s changed is all the tooling and AI on top of it. That’s the forcing function. 

NEW ACHIEVEMENT: Audiences That Belong to No Single Tool

The biggest achievement of clearing the first floor (both forms of the boss down) is that audiences stop belonging to tools.

Audiences that can be defined once, governed in one place, and deployed anywhere without rebuilding them from scratch for each platform. It’s almost like you stop paying a tax every time you want to activate a new channel or run a test.

Sarah Krasnik Bedell, the founding Growth Marketer at Railway and former Dir of Growth Marketing at Prefect, crossed over from data engineering into marketing and has seen both sides:

“The warehouse has to become the center of information because the data is scattered across too many places. Without that central hub, it’s almost impossible to get a full view of the customer or trust that data is moving cleanly across the martech stack.”

SARAH KRASNIK BEDELL, Episode 152

This new achievement is an operating model as much as an architecture: audiences defined in one place, governed consistently, available everywhere. Marketing, sales, and customer success all working from the same shared definition of the customer.

Audiences are portable, data is trustworthy, and the stack is finally coherent.

As we make our way down to the second floor, we start getting a preview of what the next battle is ahead for us, because our jobs are far from over. 

How CLI/MCP Servers Are Changing Marketing Stack Integration

CLI and MCP servers are becoming the new integration language between AI tools and marketing platforms. Activation has expanded beyond batch jobs syncing data to downstream tools on a schedule. In 2026, it’s a continuous, conversational process where AI tools read from and write back to the stack in real time. The warehouse stops being a storage layer and becomes infrastructure agents can act on.

Here’s the legendary Scott Brinker who shows up as our game guide in the next few episodes. 

“Integration has always been the hard part of martech. It was hard when we had 150 tools, and it’s even harder now with 15,000. MCP feels like a real unlock because it gives the industry a standard way for tools to talk to each other without bespoke integrations. That’s why composability is probably becoming the default architecture for martech.”

SCOTT BRINKER, Episode 201

Scott co-authored the State of Martech 2026 report, released in May 2026 alongside the annual martech landscape, and the data in it puts numbers on how fast this shift is happening.

Lots of folks are blinded by the proclamation that for the first time in 15 years, the commercial martech landscape flatlined. 

Flat is a misleading word for what’s actually happening. Underneath the near-zero net growth, the market is churning. In 2026, 1,488 products were added and 1,367 were removed. The exits are concentrated in the 2 categories that most loudly claimed the AI opportunity 2 or 3 years ago: Content Marketing had the largest outflow, with 176 products removed; Sales Automation and Enablement was 2nd. The categories gaining ground are CMS and web experience, ecommerce platforms, and iPaaS and data integration. Every growing category is benefiting from disruption hitting the rest of the stack.

Then there’s MCP. Over 29,000 MCP servers are now listed across public registries, in just 18 months. The composability that Scott’s been describing for years is now standard infrastructure.

The report argues that martech is metamorphosing from apps humans operate into infrastructure agents can use. In the last era, martech platforms competed to be where marketers worked. In the next era, they will compete to be what agents can work with.

Scott’s point gives us the map for the next room.

The first two fights were about location. The CRM had become too messy to trust as the system of record. The export reflex had made every downstream tool its own drifting copy of the customer.

MCP changes the integration layer, but it does not magically fix the thing underneath it.

Agents can talk to more tools now. Great. But if the data is messy, incomplete, stale, or missing the context behind why something happened, all we’ve done is give the machine a faster way to be wrong.

So the next floor is about the raw material itself.

One floor. One boss with 2 forms. Both down. See you on the next floor.

Episode Recap

The central argument of this episode is that the CRM lost its credibility as a source of truth through accumulation, not through any single failure. Every team that needed a record of the customer moved into the system that was already there. But folks that experienced crossing from a CRM-first org to a warehouse-first one feel a lot of friction: even when the new architecture holds more data, the loss of familiarity hits harder than anyone predicts.

The way out of this mess is 3 steps: extract raw data into a cloud warehouse, transform it inside the warehouse using something like dbt, and activate it back to downstream tools through reverse ETL. That path keeps definitions centralized, changes version-controlled, and audiences portable. Hope Barrett’s setup at SoundCloud shows what portable audiences look like at scale: everything into BigQuery as the source of truth, r-ETL sending audiences wherever they need to go, no native integrations that could introduce another drift point.

The harder argument is the Export Hydra: the habit so embedded in how marketing teams work that it doesn’t register as a choice. You build the warehouse, you centralize the data, and then you immediately start copying it outward to every tool that needs it. Each copy starts drifting. Zero-copy data sharing is what breaks that reflex, where downstream tools access a live view of warehouse data without ever receiving an export.

Many agree that the warehouse-first architecture is now the prerequisite for AI. Every agent reads from something. If what it reads is a patchwork of downstream exports, it inherits all the drift, the latency, and the competing definitions those copies carry. The warehouse-first argument won the data quality debate years ago and lost every budget conversation until AI gave it a forcing function. The underlying architecture hasn’t changed much. What changed is everything running on top of it.

Listen to the full episode for the complete playbook, including the political argument for winning the organizational battle that the technical argument alone never wins.

Full episode ⬇️ or Back to the top ⬆️

✌️

—
Cover art created with Midjourney (check out how)

Apple •Spotify• Pocket Casts •Youtube •Overcast •RSS

Related tags

See all episodes

Future-proofing the humans behind the tech

Apple •Pocket Casts•Google •Overcast •Spotify •Breaker •Castro •RSS

• LinkedIn
• Twitter
• Instagram
• YouTube
• RSS Feed