A draft appears in seconds. The campaign still takes a week to get out of the door.

That gap is where marketing leaders need to look. AI can make writing and image generation quicker, but the business only benefits when the output is useful. Someone still has to check the facts, sharpen the argument, correct the visuals and make sure everything belongs to the same campaign.

If that work grows faster than the time saved, the apparent efficiency gain becomes somebody else’s workload.

Measure the whole job

There is good evidence that AI can help with writing. In a study by MIT researchers Shakked Noy and Whitney Zhang, access to ChatGPT reduced time spent on professional writing tasks by 40%, while independently assessed quality improved by 18%.

The qualification matters. Those tasks did not require precise factual accuracy or knowledge of a company’s goals and customers. The findings show a real opportunity, but they do not establish the saving on a complete marketing campaign. MIT’s account of the study explains those limits.

Consider a hypothetical article that previously took three hours from brief to approval. AI reduces drafting from two hours to 30 minutes, but review and correction grow from one hour to two. The saving is 30 minutes. Reporting only the drafting stage would claim 90 minutes saved.

Neither figure includes a designer’s revisions or the campaign manager chasing approval. Count those too. Track hands-on effort separately from elapsed time, because a draft waiting in someone’s inbox can still delay a launch.

Bring editorial judgement in earlier

If anyone can generate another ten drafts, editorial capacity becomes a constraint. Increasing the queue does very little for the person responsible for checking it.

The answer starts before generation. Give the writer and the tool a clear audience, a commercial purpose, a point of view and source material that can support the claims. Include approved examples and explain what makes them good. ‘Make it engaging and on brand’ leaves too much open to interpretation.

Get a subject expert to challenge the angle before producing the full piece. It is easier to change an outline than to untangle a polished article built around the wrong argument.

AI can help with more than ideas. Drafting from verified notes, adapting approved copy for another channel and checking for missing information are sensible uses to test. An AI check can flag a claim that needs evidence; a named person still needs to establish whether that evidence supports it.

Give reviewers permission to reject a draft. Sometimes a fresh start is the cheaper option.

Designers inherit more than an image

A visual can look convincing in a chat window and still be unsuitable for the campaign. The product detail may be wrong, the typography inconsistent or the composition difficult to adapt for different placements. Each problem creates another decision and potentially another revision.

Then there is the volume. Imagine several people generating alternatives, each circulating their favourite. The designer now has to reconcile competing directions, identify the approved version and keep the social assets, landing page and email consistent. Generating options has created coordination work as well as creative work.

Involve design when setting the brief. Agree the visual direction, required formats and elements that must remain accurate. Use approved assets and templates where they already solve the problem. Decide which AI outputs are exploratory and which are candidates for production.

Keep one approved master, name versions consistently and consolidate feedback through one owner. Before polishing an image, ask whether it is worth taking further. A technically repairable asset may still be the wrong creative choice.

Put proportionate controls around the work

Governance should help people make decisions without turning every caption into a committee meeting. An internal concept and a public product claim need different levels of scrutiny.

A workable process needs a few things:

  • A usable brief. State the audience, objective, evidence, tone, formats and acceptance criteria. Supply approved examples alongside the instructions.
  • A named owner. Make one person accountable for the finished deliverable, with editorial, design and specialist reviewers involved where needed.
  • Checks at the right points. Approve the direction before producing variants. Check factual claims against original sources and review brand fit, accessibility and asset suitability before release.
  • Clear tool and information rules. Tell people which tools they can use, what material they can put into them and when to escalate uncertainty.
  • A controlled handover. Keep the current version, feedback and approval status in one place, so everyone knows what is ready to use.

Automate repeatable checks where they work reliably, and test them against known mistakes. Keep human attention on the judgements that affect customers, credibility and commercial value.

Train for better finished work

Start with one recurring content workflow and compare similar jobs with and without AI assistance. Record total effort, time to approval, revision rounds and the proportion accepted at first review. Assess quality against the same criteria in both groups.

Include the people doing the checking in that evaluation. A saving for the person generating content can be a cost for the person receiving it. Count tool costs and external production costs when deciding whether the change pays.

Use the results to shape training. Practise briefing, source verification, brand judgement and deciding when to stop revising. Work through real examples of acceptable and unacceptable output. Prompting is one part of that skill set.

If your team needs help, an AI Readiness and Opportunity Assessment can identify where to focus, followed by practical training around the work people actually do.

Choose one workflow, agree what good looks like and measure it through to approval. Scale the approach when the finished work justifies it.