AI prepares the marketing work; a person still decides what goes out. In XO3D ONE, completed projects feed a workflow that prepares content for several social platforms and email outreach for PR, guest posting and backlinks. Images and copy are prepared inside the application and queued for human approval. Nothing publishes without that approval.
I now manage SEO, social marketing and backlink outreach myself, without another person doing that marketing work. I spend approximately four hours a week managing it. That is my estimate of my current workload, not a measured saving against a previous baseline.
Source note: this first-person account was edited from an interview with Thomas Howcroft on 18 October 2026. It is an account of his own operating experience, not a verbatim transcript or an independently measured performance study.
What changed in my marketing role?
Before XO3D ONE, my main marketing responsibility was SEO. I wanted a clearer way to prepare the work, review it and keep the standard consistent across channels.
My role now covers SEO, social marketing and backlink outreach together. The useful change is the range of work I can manage with the system preparing material for review. My experience is that this makes consistency easier to manage. It does not establish that every post performs better, or that the system has produced a particular commercial return.
How does the AI marketing workflow work?
- Start with finished work. Completed XO3D projects provide the material for the marketing workflow.
- Prepare the content. The AI systems prepare images and copy for multiple social platforms, along with email outreach for PR, guest posting and backlink opportunities.
- Queue it for review. The material stays inside XO3D ONE for a human approver to inspect. Preparation is separate from permission to publish or send.
- Comment and revise. Reviewers leave comments in the application. AI revises the copy and updates the workflow and skill instructions to help avoid repeating the same mistakes.
- Approve before release. Nothing posts without human approval. The same boundary applies to prepared outreach: a draft is not permission to send it.
The XO3D ONE case study puts this marketing workflow in the context of the wider studio operating system.
Does reviewer feedback retrain the AI model?
No. The feedback loop changes the copy and the instructions used by the workflow. It is not model retraining. A reviewer can point out a problem, have the draft revised and make the expected standard clearer for the next preparation cycle.
That distinction matters because the human still owns the standard. An instruction update does not prove that the same error can never recur, and it does not remove the need to review the next draft.
A separate example from my work on Overshove shows what that feedback can look like:
I caught cramped layouts and cropping in videos, so we added checks for spacing and cut-off content. That feedback changes the rules across the system, not just one asset.
This is an edited quote from the interview. It describes checks added to the workflow, not a claim that every affected video has already been repaired.
What does the four-hour week actually mean?
Approximately four hours a week is my reported time managing this marketing work now. I did not measure an equivalent previous-hours baseline, so I cannot turn it into a percentage time saving or a cost-saving calculation.
I am also not claiming a verified increase in traffic, leads or revenue from this workflow. The evidence here is my firsthand account of capacity, review and consistency. Anyone evaluating a similar system should keep those operating observations separate from commercial results.
Where should a business start with AI marketing?
Start with the single most important channel for your business. For a B2B business, that may be LinkedIn. Have AI prepare the posts, then have a person approve them and comment on what needs to change before publication.
Bring the quality close to the standard you need before adding more channels. Expanding a workflow that still needs the same corrections multiplies the review work. A useful first step is one channel, a clear human approver and a repeatable way to give feedback.
This is also how I approach an AI automation pilot: start with one useful workflow, agree the approval points and evaluate the result before expanding. The aim is a dependable process that supports the person responsible for the work.