Advertising has become subsumed by marketing. That was inevitable. The moment first-party data moved into a cloud owned and governed by brands the center of gravity shifted. Demand for new customers has always been the key to everything. Now it is locked in a data warehouse.
The sweeping global data governance wave along with cookie deprecation (both real and promised) meant the data flow through the pipes of consumer attention marketplaces narrowed. First-party data you collected and owned on behalf of your customer became the new currency. Customer Data Platforms (CDPs) ushered in the era a decade ago. They consolidated first-party data and resolved identity.
The real gravity however was in the cloud. In the data warehouse. In lakes. In systems the brand controls and governs. For the enterprise this is no small matter. They understand their data is their most valuable business asset. Especially now.
Of course advertising was never the 1P data whole story. It was just the most visible piece. Marketing, properly understood, is bigger and can be better measured. Marketing is the growth engine.
Marketing includes Product Experience. Brand Value. Pricing. Loyalty. CRM. Retention. Merchandising. Channel Orchestration. Community. Advocacy. The entire lifecycle of the customer relationship from unknown to known and across touchpoints large and small. Paid, owned and earned.
Now that brands control their first-party data – and at a scale that gets bigger every passing day, advertising is being folded back into the larger operating system of marketing. One just needs to look at the Ad Agency holding companies and the services and channels they have invested in recent years namely Retail, Commerce & Influencer.
That’s the evolution we’ve just had over the past decade. So now what?
Advertising Isn’t Dead. It’s Just Noisy
There will always be advertising. There will always be supply, demand, marketplaces, exchanges. As well as automated systems for delivery, identity, pricing, measurement and optimization.
But at its core media environments have structural incentives for spend and arbitrage of dollars. Not ROI.
That doesn’t mean ad channels don’t work. They absolutely do. Search works because there is intent. Social works because there is attention. Streaming works because there is scale.
But the inefficiency is staggering. Yield optimization for publishers is weak to non-existent as is data flow from the big 3 walled-garden ad platforms (Google, Meta & Amazon). If you step back and ask, “Can we determine the optimal use of media investment?” the answer increasingly becomes: only if it’s connected to a broader system.
Which brings me to the second evolution happening now.
Marketing Has Becoming an AI Operating System
The funny thing about the questions above is “Can we determine the optimal use of media investment?” is that the answer is known to the three walled-gardens.
They know the points of diminishing returns, but you don’t. So they sell you more clicks that will not convert.
They know when you have attained optimal reach and frequency but they will make sure to burn more impressions where it increases their yield.
These are AI-driven operating systems that can be tuned for maximum efficiency of delivery and spend. But not by you, the buyer. They own and operate the media. You must have a ticket to enter and a good reason to leave if you like your job.
But what really are these platform systems? At their core these are deep learning neural networks coupled with reinforcement learning on top of a shape shifting pile of constant data streams. They learn in unsupervised, semi-supervised and supervised manners ensembling multiple models together to produce a better prediction than any single model alone.
This isn’t speculative. This is the architecture that powers:
- Amazon
Discovery + optimization.
That combination built the most powerful advertising and commerce engines in history. The question is no longer if marketing will run on this architecture. It’s whether you will build it, buy it, or keep being price gauged by the platforms based on your ignorance.
The Core Metric: CAC vs LTV
Every serious marketing system collapses down to one fundamental equation:
Customer Acquisition Cost (CAC) vs Lifetime Value (LTV).
That’s it. Every channel. Performance and Brand Advertising. Paid and owned media. Acquisition and Retention. Product, Merchandising, Pricing, Promotion, Tent-poles, Seasonality. All of what touches the customer rolls into this equation.
However if you cannot measure CAC and LTV across time and touchpoints you are flying blind. You’re running a collection of disconnected goals and campaigns.
The challenge is structural:
- LTV lives across time
- CAC lives across channels
- Revenue lives in finance
- Engagement lives in product
- Spend lives on platforms
- Identity lives in the data
It’s remarkable to me how the brands I have a direct relationship with where I am certainly in the highest tier of LTV (Levi’s, Patagonia, Spotify, Honda) act as if they don’t know me. I don;t know for certain but I would suspect none have calculated my LTV over the past decade and fed that downstream systems.
They are all smart enough to understand you can’t strategize or optimize CAC or LTV inside a single tool. You must join the data that relates in any way to these KPIs (and some you might not suspect does not). You must organize it. And it must be across time, teams and channels.
From there you can start to drive customer metrics with a closed loop system that makes you smarter and build brand equity with your customers over time by being relevant, helpful and useful to them by knowing them better.
You can move from logistic regression and tree-based models to deep neural ranking systems at enterprise scale. This means marketing is no longer a department-level problem. It’s an enterprise architecture problem. A point Andrew Rosen explored last week around the WBD deal.
AI Must Be Horizontal
I have a strong belief here. If we are going to break the chains and silos that have held us back as marketers as expressed above, AI must work horizontally.
It can’t work as “AI for marketing” disconnected from “AI for finance.” “AI for creative” disconnected from “AI for pricing.” “AI for orchestration” can’t be disconnected from “AI for bidding.”
Verticalization of AI just recreates the same issues marketers had before. Systems that are siloed and can’t work collectively as a closed loop engine.
The CAC/LTV equation touches:
- Marketing
- Finance
- Product
- Legal
- Operations
If each department runs its own isolated AI, you recreate the same data and context fragmentations that are now plaguing SaaS. You have different trains running on the same tracks without conductors. Recipe for a trainwreck.
The architectural question becomes:
Will there be one Corporate AI system to rule them all? (e.g. “Anthropic for Business”)
Or, will different AI systems be connected to share & compound their intelligence?
This is the classic bundling vs unbundling but for AI.
Do you build everything on a single foundational model provider? Or, do you compose a best-of-breed architecture?
- GPU compute where it’s optimal
- Inference where it’s optimal
- Storage where it’s optimal
- Context where it’s optimal
- Domain-specific agents where they’re optimal
Will there be an Enterprise AI application layer or does Enterprise AI become one giant service? Are we looking at a future with a myriad of future Salesforces or a few Microsofts?
My view is apps will win.
You might have:
- Marketing AI systems
- Finance AI systems
- Legal AI systems
Each with specialized applications that go deep and wide. Each connected to the other.
Data shared as:
- JSON blobs
- Structured tables
- Postgres databases
- Warehouse-native objects
The systems talk. The models reason. The organization learns.
That is the horizontal AI stack. And that is how you avoid vendor lock-in while maintaining architectural clarity.
And one note here about vendor lock-in. Because history (and memory) is so important to AI having to replace a giant service and start all your history and training over will be suicidal.
Why We Built Neuralift Close to the Data
I just don’t believe this, I’ve bet a lot on it.
At Neuralift we operate as an enterprise application directly on the brand’s first-party data connecting to their data warehouse.
That means:
- We work with first-party data.
- We respect governance and security.
- We leverage native data sharing capabilities.
- We build intrinsic value of time.
- We transform the raw data features and functions via GPU computation into business value.
Modern warehouses like Snowflake, Databricks, BigQuery and Postgres-based systems are not just storage. They are execution layers.
That’s where identity lives. That’s where revenue lives. That’s where loyalty lives. If you want to optimize CAC vs LTV, you cannot operate more than a mile away from that well. Only when you are that close to the data can you separate the signal from the noise.
From day one we decided that what we discover needs to be accessible to other AI systems as well. Data streams and downstream from us there are agents for creative builds, predictive models for next-best-action, for bidding, insight for merchandising and a map for orchestration.
VC have recently started to call this organization of data “context.” We’ve been calling it that in our product for two years. In fact, the first public demonstration of Neuralift at Snowflake Summit in 2024 showed how Neuralift works with agents (shoutout Genesis) for campaign activation. It’s nice to see that people are figuring this out now.
The AI Marketing Loop
With everything I’ve seen and heard and the work we’ve done with major brands across media, entertainment, sports, retail, travel & hospitality in a few years the self-learning AI marketing system for the enterprise will look something like this:
- First-party data + enrichments consolidated in the warehouse.
- Identity resolved at the ID level in the warehouse.
- A catalogue of tables for specific use cases containing IDs and built by agents in the warehouse.
- Neural networks training and discovering latent customer affinities by use case connected to the warehouse.
- AI interoperability enabled by open protocols.
- Multiple agents consuming that customer context and activating.
- Reinforcement learning optimizing decisions at customer touchpoints.
- Agents optimizing across systems horizontally.
- New data landing in the warehouse to close the loop and hydrate tables with new data.
Advertising is still one execution arm. CRM another. Product experience another. But intelligence is across everything. It’s not working in any one system in isolation. Not optimizing one channel at a time. Optimizing the enterprise-level objective function. CAC vs LTV.
That is the AI operating system for marketing.
The companies that win over the next decade will not be the ones with the cleverest campaigns. They will be the ones with the strongest learning loops. They will feed data into neural networks. They will let reinforcement systems optimize against real enterprise objectives. And they will ensure this marketing system can communicate horizontally across the organization to feed merchandising decisions, inventory allocation, pricing, margin forecasting and everywhere else customer data and context is helpful and will compound its value.
What’s Next?
Marketing swallowed advertising. AI will swallow marketing.
What emerges though is an operating system for the business powered by learning systems on its owned data that understand customers at the ID level as well as the overall business itself and can optimize each for long-term value and growth. One that can be tuned at any time for any objective or KPI. Just decide what use cases are most strategic this quarter.
This is in effect:
The same kind of mulit-AI system that powers Google’s Advertising.
The same kind of mulit-AI system that powers Netflix Engagement.
The same kind of mulit-AI system that powers Amazon’s Storefront.
The same kind of mulit-AI system that powers Meta’s Advertising.
Not a new invention but a new purpose. A new set of data. A new capability. One that can be built or bought piece by piece, layer by layer, optimization by optimization.
The irony is when every brand has its own walled garden, intelligence compounds as new data and new partners are added. And KPIs improve. Just like the platforms. If architectured correctly you don’t just improve marketing, you improve the entire business. That’s the evolution of marketing technology and AI.
Optimize & Prophesize 
Leave a comment