A recent conversation among marketing operations practitioners captured something many teams are quietly discovering.
A marketer had a set of HubSpot marketing emails already written in a document. The ask was simple: take the approved copy, header banner, and formatting direction, and turn them into HubSpot marketing emails. An AI browser assistant was used to help build the emails. The first attempt showed promise. When the task was mostly about applying similar content to a known template, the workflow worked reasonably well.
Then the reality of marketing operations showed up.
Updating a footer across multiple emails became a start-stop exercise. The AI needed supervision. Browser-based automation struggled in places where the email editor itself had complexity. Another practitioner suggested that starting from a template and only inserting content might be more reliable. Someone else pointed toward MCP-based approaches. Another shared a different workaround: use an AI design tool to create the HTML/template structure, then bring that output into HubSpot as a custom email template and feed the final content into it.
That exchange is useful because it moves the AI conversation away from “Can AI write email copy?” to a much more operational question:
Can AI reliably move approved marketing content into a governed marketing automation environment without breaking brand, compliance, QA, or production standards?
For most B2B teams, that is the harder problem.
The real bottleneck is not copy. It is last-mile production.
Marketing teams already have content in many places: Google Docs, briefs, campaign sheets, webinar abstracts, analyst notes, sales inputs, product messaging documents, and design folders. The challenge is not always creating more content. Often, the challenge is turning already-approved content into a live, QA-ready campaign asset inside HubSpot, Marketo, Eloqua, or Salesforce Marketing Cloud.
That last mile includes many small but critical steps:
Copy has to be placed into the right modules. Headers, preview text, buttons, body sections, legal copy, unsubscribe language, and footer fields must all land in the right place. Images need the correct dimensions and alt text. Links need tracking parameters. Personalization tokens need to be valid. Compliance fields need to remain untouched. Mobile rendering needs to be checked. The email must be associated with the right campaign, list, subscription type, workflow, and reporting structure.
This is why “AI built the email” can sound more complete than it really is.
An AI assistant may help create a draft, fill a template, or generate HTML. But the operational definition of “done” is much stricter. Done means the email is build-ready, brand-safe, compliant, measurable, testable, and connected to the right campaign architecture.
Browser agents are useful, but fragile.
The practitioner conversation also surfaced an important pattern: browser-based AI can work when the task is repetitive and the environment is simple. But email builders are not always simple environments.
Marketing automation platforms often have nested editors, drag-and-drop modules, pop-ups, preview panes, embedded frames, validation rules, and permission boundaries. A human operator understands these quirks intuitively. A browser agent may not. It can click the wrong area, lose context, fail to recognize an iframe, or stop midway because a field behaves differently than expected.
That does not mean browser agents are useless. It means they should be used for the right layer of work.
They are better for assisting with structured, repeatable steps than for owning the full production workflow. The moment the task involves compliance-sensitive edits, template logic, global footer changes, suppression rules, or personalization, human review and operations governance become non-negotiable.
MCP is promising, but it is not magic.
MCP-based workflows are exciting because they move AI from “looking at the screen” to connecting more directly with systems and data. In theory, this is the better direction. Instead of asking an AI assistant to operate a UI like a person, MCP gives AI tools a more structured way to interact with business systems.
But teams should be careful not to treat “MCP available” as the same thing as “end-to-end marketing email automation solved.”
The better question is: what exactly can the AI read, create, update, or validate in your specific HubSpot environment? Can it access the right objects? Can it update the right asset types? Does it respect permissions? Can it work with templates? Can it validate email-specific requirements? Can it handle approvals? Can it check whether the footer, subscription language, tracking, and campaign association are correct?
Until those questions are answered, MCP should be seen as an enablement layer, not a complete operating model.
The winning workflow will be template-first, not prompt-first.
The lesson for marketing operations teams is clear: do not start with a prompt. Start with the production system.
A strong AI-assisted email workflow should have five foundations.
First, a governed email template library. AI performs better when the structure is already defined. The fewer layout decisions it has to make, the lower the risk.
Second, structured content inputs. Instead of handing AI a long unstructured document, teams should provide content in a consistent format: subject line, preview text, hero copy, body section, CTA label, CTA URL, banner image, footer variation, audience, campaign name, UTM rules, and compliance notes.
Third, a design system. Brand colors, typography rules, button styles, logo usage, spacing, accessibility expectations, and image guidance should be documented in a way the AI can reference.
Fourth, platform-specific build rules. HubSpot, Marketo, Eloqua, and SFMC all have their own quirks. A generic AI workflow will break if it does not understand the platform’s template logic, editable areas, tokens, required fields, and approval process.
Fifth, a QA layer. AI should not be the final approver. It should help prepare, compare, and flag issues, but humans should still validate rendering, links, compliance, segmentation, personalization, and reporting setup.
The practical operating model
For a HubSpot email workflow, a more reliable AI-assisted process would look like this:
The campaign owner prepares the approved email content in a structured brief. The AI assistant converts that brief into a build-ready content map. The email producer uses a pre-approved HubSpot template rather than asking AI to invent the layout. If custom HTML is needed, AI can help generate the first version, but it should be reviewed against HubSpot email-template requirements and brand rules. Assets are uploaded to a defined folder. Links and UTMs are generated using a standard naming convention. A QA checklist validates copy placement, mobile rendering, footer accuracy, subscription settings, personalization tokens, image alt text, and tracking.
In this model, AI does not replace marketing operations. It reduces the manual assembly work around marketing operations.
That is an important distinction.
What RightWave believes
At RightWave, we believe the future of marketing operations is not “AI doing everything.” It is AI working inside a disciplined operating model.
For B2B teams, the real opportunity is not to let AI randomly build emails inside a browser. The opportunity is to create repeatable production loops where content, templates, data, approvals, QA, and reporting are connected.
That is where marketing operations teams can create leverage.
AI can help turn approved copy into structured email modules. It can generate HTML starting points. It can compare a live build against the source document. It can flag missing links, inconsistent footers, or off-brand language. It can help document campaign build steps. It can accelerate repetitive production work.
But without governance, it can also multiply mistakes faster.
The teams that benefit most will not be the ones experimenting with the most AI tools. They will be the ones that standardize the workflow around AI: clear inputs, controlled templates, known platform rules, human checkpoints, and measurable outputs.
The takeaway
AI-assisted email production is real, but it is still immature. The most useful insight from practitioners is not that one tool worked or another failed. It is that marketing teams are trying to solve an operational problem with tools that are still learning the operating environment.
That is why the next wave of AI in marketing automation will not be about clever prompts alone. It will be about governed production systems.
For marketing operations leaders, the question to ask is no longer:
“Which AI tool can build my email?”
The better question is:
“Have we designed an email production process that AI can safely accelerate?”
That is where the real productivity gain begins.
