Episode Details

What’s up folks, welcome to our 4 part series of Crawling through the dungeon of martech architecture. You’ve arrived at Part 2: The Eye of Context.

We cover:

• (00:00) - Intro
• (00:56) - In This Episode
• (01:28) - Sponsor GrowthLoop
• (02:32) - Sponsor: GrowthBench
• (03:32) - Welcome Back
• (04:09) - FLOOR 2: THE EYE OF CONTEXT
• (06:15) - Why AI Produces Believable Nonsense
• (09:00) - BOSS BATTLE: The Hallucination Oracle
• (10:07) - Data Quality: When Agents Read Your Messy Data
• (22:33) - Context Engineering: What It Is and Why It's Not the Same as Prompt Engineering
• (24:28) - Sponsor: MoEngage
• (25:25) - Sponsor: Knak
• (26:30) - Context Eng vs Prompt Eng
• (38:58) - Why the Industry Built the Wrong Semantic Layer in 2012
• (46:33) - How Context Rot and Fragmentation Break AI Agent Performance
• (49:59) - BOSS BATTLE: Rotten Context Mage
• (50:37) - How to Build a Shared Context Layer for AI Agents
• (58:17) - Testing Whether Your Context Layer Works
• (01:01:35) - NEW ACHIEVEMENT: The Meaning Layer Is Live

Welcome back to the Dungeon of Martech Architecture.

You’ve arrived at part 2. If this is your starting point, check out part 1 where we cleared the first floor’s boss in 2 forms: The False Truth King in the CRM, and The Export Hydra that spread it everywhere. That said, if you already have a data warehouse, you might be able to start right here.

Episode 1: CRM Gravity
We conquered the source of truth and discovered that the data warehouse replaces the CRM with portable audiences.

Episode 2: The Eye of Context
Today, we learn why AI fails without shared meaning, build the context engineering layer, and dig into why the industry built the wrong kind of meaning infrastructure in 2012.

Episode 3: The Correlation Masquerade
Next, we 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
Then, we 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.

---------------------------------------------------------------------------
FLOOR 2: THE EYE OF CONTEXT — AI Hallucinations, Data Quality, and Context Engineering
---------------------------------------------------------------------------

The layout of the second floor down the dungeon of martech architecture actually looks pretty fancy. It’s cozy, it looks modern, the whole palace is lined with mirrors. But it’s a bit creepy because once you look a little closer at the reflections, you notice that some of the details are off.

The boss on this floor is low key danger that sneaks up on way too many teams – not like the big flashy monsters from the past 2 floors.

Let’s say you have a new AI system running on your marketing data. You’ve got it producing stuff like scores, recommendations, campaign ideas. Initially, it actually looks solid.

There’s no obvious AI sentence structures in the summaries, they read well.
The scores next to accounts seem to make sense: higher ones next to well known brands and lower ones are gmail accounts.
The campaign ideas are actually pretty fresh, you can tell that it’s tailored for your ICP.
Every output is delivered with impeccable confidence.

So the next step is asking yourself… how would you know if it was wrong?

There’s a lot of obvious hallucinations that you probably catch when you chat with GPT or Claude, lik