Back to work
/ 04Applied AI

Marketing Operating System (MOS)

Compressed cross-system marketing workflow setup from days of coordination to minutes of declarative definition — and surfaced a build-on-existing alternative to a multi-million-dollar platform migration that preserves the client's existing SAP investment.

01

Days to minutes

Campaign launch compressed from 2–3 days of coordination to minutes of declarative setup.

02

Multi-$M migration deferred

Existing SAP stack orchestrated instead of replaced — capital freed for higher-priority work.

03

Running software, not slides

Working prototype in live conditions gave leadership confidence to invest.

04

AI-ready orchestration

MCP-exposed primitives let agents and operators trigger campaigns in the same grammar.

The Challenge

The client's marketing organization operated across multiple disconnected systems — campaign orchestration in Emarsys, event streaming on Kafka, downstream activation in a patchwork of point solutions. Each required its own configuration, sequencing, and human coordination per campaign.

A typical cross-system campaign required two-to-three days of human coordination across teams just to define the workflow before any creative or audience work began. A vendor proposal on the table would have replaced the existing SAP-based stack at a multi-million-dollar migration cost, betting that consolidation alone would solve the coordination problem.

The client needed a faster, more reversible path: something that could declaratively describe a campaign workflow across the existing systems without forcing a platform-replacement decision.

Our Approach

A senior applied-AI engineer designed and prototyped a Marketing Operating System — a declarative orchestration layer that sits on top of the client's existing stack.

We defined campaign workflows in a YAML domain-specific language. A marketer or operator specifies the audience trigger, the touchpoint sequence, and the activation channels; the MOS handles the cross-system orchestration. The workflow engine runs on Temporal, so campaign executions are durable, observable, and retryable across system boundaries. The event spine is the client's existing Kafka — no new data plane introduced.

We exposed the campaign primitives through Model Context Protocol (MCP), so AI agents and human operators can reason about and trigger campaign actions in the same declarative grammar. And we used the working prototype to make the case that the proposed SAP-replacement migration could be deferred (and possibly avoided entirely) by orchestrating on top of the client's existing investment rather than replacing it.

Engagement Model

One senior applied-AI engineer, full-time, multi-quarter. Reports into marketing technology and AI strategy leadership.

Technology
EmarsysKafkaTemporalMCP (Model Context Protocol)YAML DSL

Ready to get back to building?

Tell us about the engagement. We typically respond within one business day with a named builder who can talk substance — not a generic sales pitch.

Prefer email? info@inthe.cloud