Most people still use AI like a search box. Type a question, get an answer, paste it somewhere. It is quick, and the result is usually fine. Fine is the problem. Ask the same tool for a product description, a campaign idea or a report with no background and you get the average of everything it has read, which is what everyone else using it is getting too.
The work changes when you stop treating AI as a tool you pick up and start treating it as a member of the team. That means a clear job, a proper brief, the rules of the brand in writing, and a person it answers to. Most of our day-to-day marketing work runs this way now, and in our experience the gain is not only speed. The work is better too, and so is the part the people do.
What is the difference between using AI as a tool and as a team member?
A tool waits to be picked up and does one thing when asked. A team member has a role, knows the background, remembers what was agreed last time, and is expected to bring something to the table rather than just answer.
In practice the difference looks like this:
- A job, not a question. “Write me a caption” is a question. “You look after our Instagram captions: here is our voice, here are last month’s best posts, here is what we never say” is a job.
- A brief you would give a new starter. Who the customer is, what has been tried before, what good looks like, and what is off limits. AI knows a great deal about the world and nothing about your business until you tell it.
- Somewhere to keep what it learns. Brand rules, past decisions and the mistakes it made last time, written down so the next piece of work starts from there and not from scratch.
- Someone it reports to. A person who reads the work, checks it and makes the call. Nothing goes out, gets spent or gets sent without them.
None of this is technical. It is the same thing you would do for a new member of staff, and it is the step most businesses skip.
What does the research say about AI as a teammate?
The closest study to this question is a field experiment with 776 professionals at Procter & Gamble, released in 2025 as The Cybernetic Teammate by researchers from Harvard, Wharton and other schools working with P&G. People were given real product development problems to solve, either alone or in pairs (one commercial person, one R&D person), with or without AI.
People working alone with AI matched the performance of the two-person teams working without it, and spent 16.4% less time on the task than people working alone without it. Teams that had AI were around three times as likely to come up with one of the top 10% of ideas as people working alone without it. The AI also helped people outside their own speciality: without it, commercial staff leaned towards commercial ideas and R&D staff towards technical ones, but with it both produced balanced solutions. And the people using AI reported more positive feelings about the work and fewer negative ones.
Two other studies point the same way. In a study of 5,179 customer support agents, AI help raised the number of issues resolved per hour by 14% on average and by 34% for the newest and least skilled staff. The researchers found signs that the AI was passing on the best practice of the most able people. And the Boston Consulting Group experiment we covered in AI can do the work. You still have to check it found big gains on tasks the AI was suited to, and worse answers on a task it was not. So the AI needs a defined job, and a person who knows where its limits are.
Where does it bring out the best in the marketing?
More ideas get tested before one is picked
When an idea costs minutes rather than a morning, you can look at ten angles for a campaign instead of two, and throw away the eight that do not work. The final idea is still chosen by a person. It is just chosen from a better shortlist.
Someone whose only job is to pick holes
One of the most useful roles we give AI is the critic. Before a piece of work goes to a client, a separate AI is asked to find what is wrong with it: claims that are not backed up, figures that do not add up, anything off brand, anything that would be embarrassing if a customer read it. It is not polite about it, and it does not get tired of asking. A person still decides what to change, but they are deciding with the weak spots already found.
The brand stays the same everywhere
Brand guidelines tend to live in a PDF that nobody opens. Give the rules to the AI as part of its job and every caption, product description, email and advert can be checked against them each time. The tone, the words you never use, the way prices are shown, the spelling of the product names.
Nothing gets forgotten from last year
Last year’s Christmas wording, what the Black Friday banner said, which email subject line did well, which offer did not. That knowledge usually sits in one person’s inbox or memory. A team member with written notes keeps it, and brings it out when the same moment comes round again.
The dull checks happen every time
Broken links, missing alt text, a price that does not match the shop, a date in the wrong year. These are the checks people skip when they are busy. An AI with checking as part of its job can run them on every piece, and a person confirms the ones that matter.
People get time back for the part only people can do
Talking to customers. Knowing what the shop floor or the workshop is actually like. Deciding what the brand stands for, and what it would never do. Taste. The more of the groundwork the AI does, the more time goes on the things that make marketing work.
What does not change?
A team member still needs managing. AI states things with the same confidence whether they are right or wrong, so a person checks the work against the original source, by a different route from the one the AI took. We have written about that in Why AI needs a manager and AI can do the work. You still have to check it.
The decisions also stay with people. AI can prepare a campaign, a budget change or a reply to a customer. Sending, spending and publishing are done by a person who has read it first. In the UK that is not just good practice: the Competition and Markets Authority has said, in its consumer law guidance, that businesses are responsible for what an AI agent does in the same way they are responsible for what an employee does.
How do you start treating AI as a team member?
- Pick one job that repeats. Weekly social captions, product descriptions, the monthly report. Something you do often enough to see it improve.
- Write the brief once, properly. Who the customer is, your tone of voice, words you never use, what good looks like, and two or three examples of your best work.
- Keep a notes file. Every time it gets something wrong, add a line so it does not happen again. After a few weeks this file is worth more than any prompt.
- Give it a second team member to check it. A separate AI asked to find the faults, then a person who checks the facts that matter against the source.
- Decide who signs it off. One named person, and nothing goes out without them.
The short version
AI used as a search box gives you average work, quickly. AI treated as a member of the team, with a defined job, a proper brief, written brand rules, notes on what went wrong and a person it reports to, gives you better work and gives your people more time for the parts only they can do. The research so far shows AI working alongside people improves the work, and that it needs someone who knows its limits. It still needs managing, and the decisions still belong to people.
FAQ
Can AI really be a member of a marketing team?
Not in the sense of a person, and it cannot be held responsible for anything. But it can hold a defined role, work to a brief and get better at that role over time, which is how it gets the most useful results. Treating it as a team member is about how you set it up and manage it.
Will AI replace my marketing team?
The evidence so far points to AI making people more effective rather than replacing them, especially less experienced staff. What it changes is how the time is spent: less on groundwork, more on decisions, customers and ideas.
What marketing jobs should AI take on first?
Jobs that repeat and have a clear standard: captions, product descriptions, first drafts of emails, pulling figures together for a report, and checking work against brand rules. Leave the decisions, the spending and anything sent to customers with a person.
How do I brief AI like a team member?
The same way you would brief a new starter. Tell it who the customer is, what the brand sounds like, what it should never say, what has been tried before and what good looks like, with real examples. Keep the brief written down and update it every time something goes wrong.
Is this only for big companies?
No. A small business can do all of this with the AI tools it already pays for. The brief, the notes file and a named person who signs the work off cost time, not money.
Sources: Dell’Acqua et al., The Cybernetic Teammate: A Field Experiment on Generative AI Reshaping Teamwork and Expertise, NBER Working Paper 33641 (April 2025); Brynjolfsson, Li and Raymond, Generative AI at Work, NBER Working Paper 31161 (2023); Dell’Acqua et al., Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality, Harvard Business School Working Paper 24-013 (2023); CMA, Complying with consumer law when using AI agents (March 2026).