Beyond Prompts: How AI Is Evolving from Content Generator to Email Operating System

We need to shift our perspective from mere content generation to an operating system. In this article, we’ll look at AI on different levels (copywriter, orchestrator, and strategist), highlight a missing layer between user intent and a production-ready email, and clarify where your team stands and what you need to move forward.

Level 1. AI as copywriter

This is where many companies begin: brainstorming, subject-line generation, copywriting, and adapting emails for different audiences, you name it. It’s a beginner level, but it brings value to teams. At this level, it is important for the organization to have streamlined processes and workflow to allow AI to make decisions within a framework and guardrails set by the organization.

HUMAI’s work with FMCG brands is a good example. Their creative CRM workflow uses AI across briefing, copy creation, asset creation, and email development. Of course, humans still control the output at each step. This approach saves time and costs: according to HUMAI, it can cut briefing and content creation time by up to 70 percent.

But there are some limitations:

• Lack of context: AI doesn’t know your brand library, design tokens, or available sections and generates them from scratch every time.
• Brand consistency depends on humans: Without governed rules, AI can drift from guidelines. The human reviewer is the only quality gate.
• The “email assembly” step: You still need to manually assemble generated content into a production-ready email.
• No learning loop: Campaign results don’t feed back into the system. They stay in an email marketer’s memory instead of a knowledge base.

So, yes, this first level is useful, but it’s still local optimization. It speeds up separate steps without changing the email production architecture.

The missing layer

Here’s the catch: The problem isn’t content generation. The issue is the absence of a semantic layer between user intent and production-ready email. Without it, AI is kind of blind. Yes, it will generate your copy and images, but it doesn’t know which email parts are unchangeable brand assets, which it can edit, or how to put together an email for different segments. This prevents scalability—and that’s where the next level begins.

Level 2. AI as orchestrator

The shift to level 2 needs a governed composition. Here, AI doesn’t improvise an email structure but works within a semantic system. At Stripo, this is the idea behind the Email Domain System (EDS), a semantic composition framework built on top of Stripo’s editor infrastructure. It treats an email as a composition of semantic elements, each with a role, content slots, and controlled variants.

The pipeline looks something like this:

Intent → Plan → Composition → Rendering

• Intent: Parse the campaign goal, audience, and references.
• Plan: Build a list of semantic sections with roles, variants, and rules (we’ll call it a brief).
• Composition: Choose components, apply decorators, and fill content into typed slots.
• Rendering: Transform the semantic structure into production-ready HTML.

Note that the brief is important here. It’s a semantic plan that defines which sections your email needs, their order, and the data they require.

Here’s what you might specify in your brief for a back-to-school campaign:

• Hero banner for attention and framing
• Product showcase for the main offer
• Benefits block for proof
• Countdown timer for FOMO
• Footer for compliance

Brand consistency by construction

At Level 1, people check the design, content, and layout. At Level 2, the architecture itself helps us stay consistent. Governed tokens handle colors, typography, and spacing. Atoms are the smallest semantic units (e.g., for CTAs, titles, and prices). Components combine these building blocks into structures. Variant swap lets the team change the layout without rebuilding the content.

This means A/B testing can become a configuration change instead of a redesign. With one brief, one set of content, and two variant overrides, you get two distinct emails. Tone of voice is also part of your system. It’s like a guardrail for generated copy that keeps the brand sounding like itself, regardless of which model or prompt you used. Human validation isn’t eliminated; it becomes more structured and less repetitive.

Interactive content as a new dimension

At this second level, you can also move from static emails to new content types. Widgets open the door to interactive content, a new type of email module in which content is configured through natural-language conversation with an AI assistant.

You just describe what you want (e.g., “a Wheel of Fortune with a seasonal promo for dormant subscribers”), AI turns it into a structured configuration, and a microservice renders it into email-safe HTML.

This gives you access to interactive elements like polls, questionnaires, quizzes, and scratch cards that are hard to create and scale manually, all without requiring an email or coding expertise. Rather than creating another static campaign under time pressure, you can use everything that email has to offer.

Level 3: AI as strategist

This is the level at which the system becomes autonomous and starts making decisions, still with human guidance. For organisations this implies clear roles, processes and governance models to ensure AI gets to reach it’s maximum potential with’human-in-the-loop’ roles clarified. For instance, a person defines a business goal (“increase retention by 20 percent”), and AI determines what to send, to whom, through which channel, with what content, and at what frequency. It makes these choices based on goals, feedback, and accumulated knowledge.

This isn’t science fiction, but it’s where many teams are not yet operating. It requires both agentic-ready organisation, as well as an Experiment Knowledge Base (EKB) that stores A/B test results with hypotheses and outcomes, audience segments, content rules, brief modifiers, campaign strategy presets, and a feedback loop. That loop closes the cycle:

Brief → Email → Send → Metrics → Knowledge → Next Brief

Here, the human role shifts from operator to architect and oversight. Instead of checking each output, the focus is on defining rules and guardrails that make results reliable. It’s clear that “AI helps the organization improve its email marketing over time” is a better outcome than “AI generates content for us.”

How does it work in practice?

Here’s how companies at different levels of AI use can approach the same campaign:

Final thoughts: How to start shifting

It’s important to understand where your team is now. Even if you see yourself at Level 1, that’s perfectly fine. It’s not like you need a platform overhaul. Instead, take three steps:

1. Ensure ‘agentic readiness’ in the organization. Involve the wider organization in the agentic workflow and define and clarify roles and responsibilities across the process from ideation to launch and measurement. Define the roles for both humans and AI across the development and delivery process.
2. Define your brand semantically. Create a profile covering your brand, tone of voice, and design tokens to “feed” AI.
3. Move from templates to component systems. Build governed semantic components with controlled variants.
4. Gather experiment data. Even if you don’t have a feedback loop yet, document A/B testing results with hypotheses and results to build the knowledge base for Level 3.

When you treat AI as a junior copywriter, you can save time on individual tasks. But building the semantic infrastructure for AI to orchestrate will change how email marketing operates. The technology is ready. Is your architecture?

This is an article by EMAS 2026 Gold Sponsor HUMAI/Stripo.

Curious how to put this into practice? Visit EMAS 2026 on June 11 at Circa Amsterdam. Be inspired by international speakers and discover how to take your email marketing to the next level. Tickets and programme: emas.nu

Dmytro Kudrenko

Founder and CEO | Stripo

• LinkedIn