Every software update now comes with an AI button, and every conference talk promises AI will change your marketing. Most SME owners we speak to feel they should be doing more with it, without being sure what “more” means.
What decides whether AI pays off is where you use it. Put it on the right jobs and it saves hours every week. Put it on the wrong ones and you get bland copy and wrong numbers, with nobody accountable for either.
This month we published a statement for our clients setting out where we use AI in their work and where we don’t. This post is built on it.
Where AI fits
AI is strongest on work that is repetitive, heavy on inputs and easy to check.
Research and analysis. Summarising sources, scanning competitors and markets, and finding themes across large amounts of text. When Google announced new product fields for AI-powered Shopping results, Mary on our paid media team used AI to pull the research together. She estimates it would have taken over a week by hand. It took a day.
Drafting. First drafts of copy, subject lines, ad variations and briefs. Search ads need plenty of headlines, and AI can produce a long list of angles for a person to shortlist and check against ad policy. Someone on the team reviews and improves every draft before it goes anywhere.
Creative production support. Layout ideas, resizing one design into every format a campaign needs, and producing variants for testing.
Data and reporting. Cleaning and structuring data, building reports and drafting the commentary, which a person then checks against the source. Jason and Mary on our PPC team were spending around 3 to 4 hours a week between them keeping recurring reports up to date, which is why reporting was one of the first jobs we started automating.
Internal operations. Meeting notes, task summaries and internal documentation.
Where it doesn’t fit (yet)
Strategy and budget decisions. Recommendations should come from people who know the business and its market. AI can research a decision and summarise what your competitors say. Deciding what your business stands for stays with you.
Final approval. A named person should review and sign off anything before it reaches a customer. Our client statement says it plainly: a person is accountable for every deliverable, and AI does not approve or sign off work.
Acting on its own. Nothing should be sent, published or spent without a person approving it. AI tools that can post, email or change budgets directly are useful, as long as a person approves each action.
Facts nobody has checked. Prices, dimensions, stock levels, lead times and warranty terms are exactly where AI goes wrong. It fills gaps instead of admitting it doesn’t know, and it does so with total confidence. AI meeting notes can mishear a number or flip a negative, so “we won’t be doing X” arrives as “we will be doing X”. Check anything factual against the source before it goes live.
Your brand voice. AI writes first drafts with a house style of its own, and customers have got quick at spotting it. Once they do, the page stops sounding like your business. Show the tool four or five pieces of your writing rather than describing your tone, and read the final version aloud. If you wouldn’t say it to a customer standing in front of you, it needs another pass.
A simple framework: Automate, Assist, Keep human
We sort marketing work into three groups.
| What it means | Examples | |
|---|---|---|
| Automate | AI does the task end to end, with regular spot checks | Moving data between systems, updating recurring reports, meeting notes to tasks, resizing images |
| Assist | AI produces a draft or analysis, and a person edits, fact checks and approves it | Research briefs, first drafts, ad copy variations, performance commentary |
| Keep human | People decide, and AI can inform | Strategy, budgets, brand positioning, final sign-off, anything sent or spent on your behalf |
To use it, list your team’s regular marketing tasks and put each one in a column. Be strict about the Automate column. Anything a customer will see usually belongs in Assist, however good the tool looks.
Common mistakes
Starting with the tool. Buying a platform and then looking for a use for it gets things backwards. Start with the task that wastes the most time, then find a tool that fits.
No guardrails. Agree which tools are approved, what data can go into them and who checks the output. Keep customer names, addresses and order data out of free or personal AI accounts. We use business tier accounts set up so that client content isn’t used to train the providers’ models.
No measurement. If you don’t know how long a task took before, you can’t tell whether AI helped. Log the time before and after, and track errors and rework as well as speed.
Hear the full session live on 13 October
We’ll go deeper on this in Qoob’s marketing session at Accelerating AI, a free conference on Tuesday 13 October at Unity Place, Milton Keynes. Lunch is included, and speakers include Canva, Microsoft, NatWest and FreeAgent.
Book your free place or visit acceleratingai.co.uk.
For five practical examples of AI in marketing, read 5 Ways SMEs Are Using AI in Marketing Right Now