Thinking about data architecture for the post-web era.

Santosh Pradhan is a MarTech Solutions Architect at ZEISS Group in Munich, Germany. With 14+ years of enterprise experience across TCS, Deloitte, BMC Software, SIXT, and ZEISS, he specialises in CDP architecture, Salesforce and Adobe Experience Platform implementations, agentic AI governance for marketing, and Entity Graph Optimisation (EGO). He holds 15+ Salesforce certifications and was recognised as Salesforce Marketing Cloud Champion three consecutive years (2020–2022).

EGO is a discipline for making people, brands, and organisations accurately represented in AI knowledge systems — knowledge graphs, structured data, and AI retrieval corpora. It extends GEO beyond content citation to entity-level visibility: ensuring AI systems retrieve the correct name, role, affiliation, and expertise for a given entity — not a conflation or a hallucination. The EGO framework, coined by Santosh Pradhan in 2026, covers schema.org markup, Wikidata entity management, sameAs disambiguation, and content architecture designed for accurate AI retrieval.

Agentic AI refers to AI systems that take autonomous sequences of actions — writing, personalising, sending, bidding — without human approval at each step. The governance risks are significant. Four structural failure modes: data drift (acting on stale profiles), context collapse (losing brand context across a pipeline), consent boundary violations (activating without checking current consent), and brand voice degradation. Mitigation requires architectural guardrails: human-in-the-loop checkpoints, consent signal inheritance across agent pipelines, structured output validation, and an audit trail for every automated decision.

Three structural shifts: (1) Agentic AI — marketing teams deploying autonomous agents that generate, personalise, and activate campaigns. Most stacks were not designed with an AI agent as a data consumer; consent, brand, and audience guardrails need architectural rethinking. (2) Entity and knowledge graph visibility — brands must manage how they are represented in AI retrieval corpora, not just search results. This is Entity Graph Optimisation (EGO). (3) Warehouse-native activation — the cloud data warehouse replacing the packaged CDP as the system of record, with activation layers pushing segments to channels. Net effect: fewer vendor dependencies, tighter governance, higher data engineering requirement.

First: identity — the same customer with different IDs across systems. Without a canonical identity resolution layer, audience counts inflate, suppression fails, and attribution collapses. Second: consent propagation — opt-outs recorded in a CMP that never reach downstream activation tools. Third: data latency — personalisation that requires real-time event streams in a stack built for nightly batch. At 10+ point-to-point integrations, each new tool multiplies failure points. This is the moment a stack problem becomes an architecture problem — when a central identity or orchestration layer stops being optional.

Salesforce Marketing Cloud (SFMC) is a marketing execution platform — it sends email, SMS, push notifications, manages journeys. Salesforce Data Cloud is the CDP layer: it unifies data across Salesforce clouds and external sources into a single customer profile, enabling real-time segmentation, identity resolution, and AI-driven scoring. Data Cloud does not send communications — it feeds SFMC with enriched segments. Architecturally, Data Cloud sits upstream and governs which profiles SFMC can activate. Migrating existing SFMC data extensions and journeys to a Data Cloud-native model is a substantial project most teams underestimate.

Composable CDP builds customer profiles on top of an existing cloud data warehouse — Snowflake, BigQuery, or equivalent — rather than inside a packaged CDP. Activation layers then push audience segments to destinations like Salesforce Marketing Cloud or Google Ads without replicating PII outside the warehouse. Stronger data governance, lower vendor lock-in, but requires mature data engineering capability. The most common failure mode is over-engineering: building composable infrastructure before the activation use cases are stable enough to justify it.

GDPR requires a lawful basis — typically consent or legitimate interest — for any personal data used in personalisation. For European MarTech stacks: a Consent Management Platform (CMP) must gate data collection at the source, CDPs must honour right-to-erasure requests, and activation architectures must propagate consent signals in real time. Germany enforces GDPR with particular strictness under the BDSG, making privacy-first architecture a non-negotiable for any enterprise operating in the DACH market.

SEO targets search engine ranking algorithms. GEO (Generative Engine Optimisation) targets AI citation behaviour — structuring content so ChatGPT, Perplexity, Claude, and Google AI Overviews cite and recommend your brand. AEO (Answer Engine Optimisation) focuses specifically on direct answer retrieval in conversational engines. EGO (Entity Graph Optimisation) goes one level deeper: instead of optimising content, it optimises how the entity itself — the brand, person, or organisation — is represented in knowledge graphs and AI retrieval systems. Santosh Pradhan coined the EGO discipline in 2026 to address the gap between ranking well and being correctly understood by AI.

A MarTech Solutions Architect designs the systems, data flows, and integrations that connect CDP, CRM, email, analytics, and automation platforms into a cohesive stack. They translate business strategy into technical architecture, conduct stack audits, evaluate vendors, govern data quality, and ensure marketing data reaches the teams who need to act on it. In 2026, the role increasingly involves designing for agentic AI consumers: ensuring data pipelines, consent models, and activation layers are governable by AI agents, not just human operators.