The Proprietary Intelligence – 1
Before defining what Meridian is, it is worth being precise about what it is not. The market is saturated with “AI-powered marketing” claims. Most of them describe incremental improvements to existing approaches. Meridian describes something structurally different.
Meridian is not a generic large language model wrapped in a marketing interface. Generic LLMs are trained on public internet text. They can write. They can summarise. They can generate variations. What they cannot do is reason about your specific customers, your specific products, your specific margin structures, your specific competitive dynamics, your specific history of what worked and what failed. Wrapping a general-purpose AI in a marketing user interface produces general-purpose results — which is to say, mediocre results that any competitor can replicate with the same wrapper.
Meridian is not journey automation with better copy. Most “AI-powered” marketing tools take the existing campaign-mode paradigm and accelerate it: more segments, more variations, more journeys, more tests. They optimise within the existing model without questioning whether the model itself is the problem. Faster mediocrity is still mediocrity. More sophisticated segmentation is still segmentation. If the underlying approach loses eighty percent of engaged customers per quarter, making that approach faster does not solve the problem.
Meridian is not a dashboard or insights layer. The industry has no shortage of analytics that tell marketers what happened. Dashboards are excellent at describing the past. They are useless at changing the future. Insights without action create sophisticated paralysis — teams that know exactly how they are failing but lack the operational capacity to do anything about it. Meridian does not primarily report. Meridian acts.
So what is Meridian?
Meridian is a proprietary decisioning model built on Context Graphs that creates a BrandTwin for each valuable customer, orchestrated by agents that act continuously and autonomously, and priced on the outcomes it delivers.
That definition has four layers. Each layer matters. Each layer builds on the one before.
Layer 1: Context Graphs as Substrate
[See Context Graphs: The Missing Layer Between Data and Action
2026 AI and Marketing and NEO’s Context Graph Advantage: Turning Customer Drift into Compounding Attention]
Most customer systems store data. Meridian requires something different: structured intelligence optimised for reasoning and action.
The distinction matters. A traditional customer data platform stores facts: this customer bought this product on this date, opened this email at this time, visited this page in this session. Facts are necessary but insufficient. Facts tell you what happened. They do not tell you what it means or what to do about it.
A Context Graph is not a dashboard. It is not a data warehouse. It is not merely a knowledge graph with marketing data. A knowledge graph answers the question: what do we know? A Context Graph answers a different question: what matters now, and why?
The architecture has three interlocking components.
The Customer Context Graph captures state and trajectory. It knows not just where a customer is but where they are heading. This includes identity and relationship history across channels, preferences both explicit and inferred, engagement rhythms and channel behaviours, lifecycle stage signals, value trajectory including lifetime value and contribution margin and churn risk, interaction history across messages and sessions and transactions, and crucially, confidence scores that quantify how certain the system is about each inference. Not all knowledge is equal. The system must know what it knows and what it merely guesses.
The Product Context Graph captures relationships and economics. It understands catalogue structure — which products are substitutes, which are complements, which are accessories, which are gateways to larger purchases. It tracks inventory and availability in real time. It knows margin and unit economics at the SKU level. It monitors return rates and dissatisfaction signals. It understands seasonality and demand patterns. This layer ensures that recommendations are not just relevant to the customer but profitable to the business and actually available to fulfil.
The Decision Trace Graph captures what most systems entirely ignore: why decisions were made. Not just that a discount was offered, but what signals triggered the intervention, what alternatives were considered, what hypothesis was being tested, what the outcome was, and what precedent was established for future similar situations.
This is the critical differentiator. Most enterprise software manages the “state clock” well — what is true right now. Almost nothing manages the “event clock” — what happened, in what order, with what reasoning, producing what result. When humans were the decision layer, the “why” lived in people’s heads. Institutional knowledge was reconstructed through conversation and experience. But AI agents do not have heads. They cannot remember why a decision was made unless that reasoning is explicitly stored. Without decision traces, every choice starts from scratch. The system cannot learn from its own experience. It cannot transfer insights from one customer to another. It cannot compound intelligence over time.
Meridian’s Context Graph is built to hold both state and reason. The state tells you where you are. The reason tells you how you got there and what to do next.
