AI-assisted content workflow with human approval
An automated weekly workflow that proposes topics, drafts posts and creates images with AI, with a person approving every post before it is published.
- Client
- Senior leader, pharmaceutical sector
- Services
- AI automation, workflow design
- Area
- Development and automation
Context
The wider situation.
Senior leaders are increasingly expected to share their perspective publicly, and regular thought-leadership content can help them engage with peers, partners and their wider sector. Producing it consistently, however, takes time. Choosing relevant topics, writing posts and preparing images competes with the leader's main responsibilities, and content plans are often the first thing to slip when other pressing priorities arise.
AI tools can now propose topics, draft text and create images, which makes regular publishing more achievable. They also introduce a risk. Content published under a person's name reflects their judgement and reputation, and in sectors such as pharmaceuticals there can be additional sensitivity about what is said publicly. Automating the whole process would remove the editorial control the author needs, so the design question is where AI helps and where a person decides.
The challenge
What needed to change.
Producing regular thought-leadership content competed with the leader's other responsibilities.
Handing it to AI alone would risk losing editorial control over what is published under their name.
What we did
Our approach.
Designed a weekly workflow that proposes relevant topics
Used AI to draft posts and create images
Routed every post to a person for approval before publication
Kept the decision on what is published, and when, with a person
The workflow was designed around a single principle: AI supports drafting, while a person decides what is published. It runs on a weekly cycle that proposes relevant topics. AI is used to draft posts and create images, and every post is then routed to a person for approval before publication. The decision on what is published, and when, stays with a person throughout, so the leader keeps editorial control over content that appears under their name while the preparation work is automated.
Delivery perspective
AI supports drafting, while the person publishing the content retains editorial approval.
Planning similar work
What similar organisations should consider.
Decide where people stay in control
Before automating any part of content production, decide which decisions must remain with a person. Approval of what is published, and when, is the most important of these. Designing the workflow around that decision ensures that AI speeds up preparation without changing who is accountable for the final content.
Use a regular rhythm
A fixed schedule, such as a weekly cycle, gives the workflow a predictable pattern. The author knows when proposals will arrive and when approval is needed, which makes it easier to fit review into a busy diary. A steady rhythm also tends to produce more consistent publishing than occasional bursts of activity.
Treat AI output as a draft
AI-generated topics, posts and images should be treated as starting points. They may need editing for accuracy, tone or relevance to current events. Positioning them clearly as drafts sets the right expectation for the reviewer and encourages careful reading rather than routine approval of whatever the workflow produces.
Make approval quick and clear
If approving a post is slow or awkward, the workflow will either stall or be bypassed. Routing each post to the approver in a simple form, with the draft and image together, makes review practical. A clear approve or decline step also keeps a record of what was published and on whose decision.
Discuss a similar project.
Tell us about your project.