Embedded-or-Vertical AI Distribution Rule
Embed broad AI into existing habits; make focused AI a destination.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 98%
Classify an AI experience by the breadth of the job it performs. Broad, general-purpose capabilities resemble search: users benefit when they appear inside applications and workflows they already use, rather than requiring another destination and habit. A narrow vertical experience can justify a dedicated destination because users arrive with a specific intent and can clearly understand why they should return. The mechanism links use-case breadth to distribution: horizontal capability plus existing workflow leads to embedding, while focused capability plus strong intent can support a standalone product. Teams should test the strength and frequency of that intent before committing to destination-level acquisition costs.
Origin
Extracted from Marketing Against The Grain during a comparison of Snapchat My AI, ChatGPT, and Character.AI.
Core principles
- 01Broad capabilities benefit from existing distribution.
- 02Focused outcomes can sustain dedicated destinations.
- 03Habit formation is harder for undifferentiated horizontal tools.
- 04The specificity of the use case should determine the delivery model.
How to run it
- 1
Define the user job
State the exact task the AI performs and the person performing it. Separate a broad conversational capability from a focused outcome.
Pro tip Write the job as a verb-object statement, such as diagnose a symptom or draft a campaign.
Watch out Do not classify the product merely by the model or interface it uses.
- 2
Measure use-case breadth
Determine whether the capability serves many unrelated intentions or one coherent vertical workflow.
Pro tip Review actual prompts or proposed use cases for evidence of fragmentation.
Watch out A long feature list can disguise the absence of a focused job.
- 3
Map existing destinations
Identify applications or workflows where the audience already performs the task. Prefer embedding when those destinations own the relevant habit.
Pro tip Look for integration points immediately before or after the AI-assisted task.
Watch out Creating a new destination also requires creating a new recurring behavior.
- 4
Test destination intent
Ask whether users would deliberately visit and possibly pay for this specific experience. Strong, repeatable intent supports a standalone vertical product.
Pro tip Validate return behavior rather than relying only on initial curiosity.
Watch out Novelty-driven traffic can look like durable destination demand.
- 5
Choose and validate distribution
Embed horizontal capabilities or launch a focused vertical destination, then measure recurring usage and retention.
Pro tip Compare embedded feature adoption with standalone return frequency where feasible.
Watch out Do not preserve the original distribution model when behavioral evidence contradicts it.
In the wild
Snapchat adds a simplified conversational assistant inside a destination its young audience already uses. Users receive short answers and advice without establishing a separate ChatGPT habit.
→ Broad AI gains adoption through Snapchat's existing distribution and behavior loops.
Character.AI offers a specific experience: interacting with simulated celebrities, authors, and fictional or historical figures. The focused promise gives users a reason to visit the product directly and supports a paid tier.
→ A vertical lifestyle use case sustains destination-level engagement.
Common mistakes
Making horizontal chat another destination
A general chatbot must create a new habit while competing with stronger horizontal platforms. Embed it where the task already happens instead.
Embedding a high-intent vertical too deeply
A valuable specialist workflow may deserve its own destination, brand, and payment model rather than remaining an invisible feature.
Is it for you?
Best for
Product leaders and founders choosing the distribution model for a new AI experience.
Not ideal for
Products whose audience, recurring task, and value proposition remain undefined.
From the transcript
“My argument is those are best as embedded features.”
“Those are going to be the destination AI apps where you're actually gonna go directly to them”
“Much easier to build AI chat, integrate it into an existing destination or to build it in a vertical way and solve a specific problem…”
From the episode
Why Snap MyAI is the Sleeping Giant of AI Chat (#148)