Jon Miller's Recent LinkedIn Posts

Here’s what I’d add to Scott Brinker’s (excellent) research on the New Martech “Stack“ for the AI Age. BRINKER'S THESIS Brinker argues the traditional martech stack is dissolving into a composable canvas around a unified data platform. Five classes of data (Customer, Company, Content, Code, and Control) converge on one shared foundation, with apps and agents composing dynamically on top. He organizes this into five rings: A Data Core and Semantic Layer at center, Context-as-a-Service (CaaS) and Decisioning in the middle, Apps and Agents at the edge. WHAT DOES THIS MEAN FOR MARKETING AUTOMATION? MAPs have historically been the core of the martech stack. In Brinker's framework, the outer ring is commoditized delivery: receiving audiences, sending emails. The CaaS and Decisioning rings are where valuable work happens: orchestrating actions while understanding context and applying governance. I think this points to marketing automation’s future as a marketing-specific CaaS and Decisioning service. Not a system of record for data (that's the data platform), but for marketing context: what campaigns exist, who's eligible, what rules apply, what you've learned about what works. Context built through actual campaign execution (which is much more than sending emails). The data platform knows what you know. The marketing context platform knows what you're doing about it. FOUR TYPES OF MARKETING CONTEXT Here are four types of context for AI-native decisioning.   1️⃣ Operational guidance. Which fields to use and when, naming conventions, UTM standards. Think claude.md in Claude Code: it’s the operations and customer guidance you’d give a new MOps hire, referenced by AI every time it acts. 2️⃣ Memory. Patterns learned by observing your team and customers: case studies outperform whitepapers for enterprise, Tuesday sends convert better for webinars. This compounds over time, like Claude builds memory across conversations. 3️⃣ Governance. Frequency limits, consent rules, suppression logic. These are hard, structured rules, not freeform guidance. 4️⃣ Brand and voice. Tone, terminology, and style so AI-drafted content sounds like your company. The risk is this becomes another accidental silo. Every mature Marketo instance has years of institutional knowledge locked inside: campaign logic, field preferences, scoring models. Valuable, but trapped in spaghetti that can't be queried by AI or composed with other tools.  AI-native context needs to be accessible and cross-referenced across systems, not locked into one platform. WHERE THIS LEADS As AI agents proliferate, every agent needs marketing context to avoid "confident errors", e.g. technically correct actions that are strategically wrong. The MAP becomes a marketing context and orchestration service, built on structured data instead of spaghetti, queryable via open protocols like MCP, and compounding institutional knowledge through usage. Is this a new category, or the old one rearchitected?