If you run marketing on your own, you have been told that AI will give you your week back. Sometimes it does. Sometimes you spend forty minutes coaxing a tool into producing something you could have written in thirty.
The difference is mostly about which tasks you hand over. This guide sorts the common ones into three piles and gives you a way to sort the rest yourself.
The test: cost of error and ease of checking
Two questions decide which pile a task belongs in.
How bad is it if this is wrong? A clumsy internal summary costs you nothing. A wrong price on a landing page costs you money and an awkward phone call.
How quickly can you tell whether it is right? You can judge a list of headline options in seconds. Checking a 2,000-word technical article properly takes almost as long as writing it.
Low cost of error and quick to check: hand it over. High cost or slow to check: the tool can assist, but you do the finishing. If the value of the task lies in your own judgement or relationships, keep it.
The three piles at a glance
| Task | Pile | Why |
|---|---|---|
| Meeting notes and action lists | Hand to AI | Low stakes, quick to skim for errors |
| Summarising research and long documents | Hand to AI | You can spot-check against the source |
| Repurposing a finished piece into other formats | Hand to AI | The thinking is done. This is reshaping |
| Alt text for images | Hand to AI | Fast to check, tedious to write |
| Tidying spreadsheets and contact lists | Hand to AI | Rule-based. Verify a sample |
| Headline and subject line options | Hand to AI | You choose. It only supplies candidates |
| First drafts of articles and emails | AI drafts, you finish | Structure is cheap. Substance and voice are yours |
| Monthly report commentary | AI drafts, you finish | Numbers must be checked. Meaning is your call |
| Social posts | AI drafts, you finish | Easy to make bland. Needs your voice |
| Keyword and topic research | AI drafts, you finish | Good for ideas. Verify with real search data |
| Positioning and messaging | Keep human | This is the strategy itself |
| Customer interviews | Keep human | The relationship is the point |
| Final copy and sign-off | Keep human | Your name is on it |
| Anything legal, regulated or factual | Keep human | High cost of error |
Pile one: hand to AI today
These are the dependable savings. None of them is glamorous.
Meeting notes. Record the call (with everyone’s agreement), let a tool transcribe and summarise it, and skim the action list while your memory is fresh. Correct anything it misheard, especially names and numbers.
Research synthesis. Paste in five long articles or a stack of survey comments and ask for the common themes, the disagreements and what is missing. Treat the output as a map of the material, then read the parts that matter yourself.
Repurposing. Once a piece has been written and approved, turning it into a newsletter item, three social posts and a short video script is reformatting. A tool does this well, because the facts and opinions are already fixed.
Alt text. Useful for accessibility and search, dull to write. Generate it, glance at it, correct anything that misdescribes the image.
Data tidying. Standardising job titles, splitting names, removing duplicates, grouping free-text responses. Describe the rule, check a sample of the results. Keep personal data out unless your AI policy allows that tool for that purpose.
Options. Twenty headlines, ten subject lines, five ways to structure a talk. You are choosing, not accepting.
Pile two: AI drafts, you finish
This is where most of the over-promising happens. The tool gets you to a first draft quickly, and a first draft is perhaps a third of the work.
Articles and emails. A draft from a good brief saves you the blank page. It will not contain what only you know, and it will sound like everyone else until you edit it. Plan for the edit. See how to brief AI and humans for a brief that produces a better starting point.
Reporting. Tools are good at turning a table of figures into paragraphs. They are also capable of describing a trend that is not in the data. Check every number against the source, then write the “so what” yourself. Your CEO wants your reading of the numbers, not a description of them.
Social posts. Fine for volume, weak on personality. Draft a batch, keep half, rewrite those.
Keyword and topic ideas. A model can suggest what people might ask about a subject. It does not know actual search volumes unless it is connected to a data source. Confirm ideas against Google Search Console or a proper keyword tool before you build a plan on them.
Pile three: keep human
Positioning. Deciding who you are for and why they should choose you is the job. A model can help you pressure-test a position. It should not pick one, because it will pick the average, as explained in why AI makes every brand sound the same.
Customer interviews. The value is in the pause before an answer, the thing they nearly did not say, and the fact that somebody from your company bothered to ring. Use AI to summarise the transcript afterwards.
Final copy. The last read, the last cut, the decision to publish. If a person has not approved it, it is not finished.
Anything legal or factual. Terms, compliance statements, product specifications, pricing, claims about results, anything about health, safety or money. A tool can help with layout. The content comes from the person who is accountable for it.
Do the sums honestly
A saving is the time the task used to take, minus the time it takes now including prompting and checking.
Say a blog post takes you four hours from scratch. With a good brief, an AI draft and a proper edit, it might take two and a half. That is a real saving, and it is nowhere near the “four hours down to ten minutes” you will hear in a sales demo. The ten-minute version exists. It is the one your readers scroll past, and this is how they spot it.
Time three or four of your own recurring tasks before and after. Those are the figures to give your CEO, because they are yours.
Where to start this week
- List everything you did last week, roughly, with time against each item.
- Mark each with a pile: hand over, draft and finish, keep.
- Pick the two largest items in the first pile and move them to a tool this week.
- Pick one from the second pile and time it properly, including the edit.
- Decide what you will do with the hours you get back, before they fill up with something else.
Step five matters more than it looks. Time saved on notes and reformatting is only worth having if it goes into the third pile: the customer calls and positioning work that never get done. If you want a structure for that, a 90-day marketing plan for a team of one is a good next read, and the tool stack worth having covers what to use.
There will also be things in pile two you simply do not have time to finish properly. That is the gap we fill at DGTL: AI-assisted work with a named person checking and signing it off, so the finishing still happens.