Nearly half swapped tasks for AI. Most can't say whether it paid off
Before you hand a task to AI, write down what it costs you today: the hours it takes, how long it sits between brief and delivery, what each piece of work costs. Sooner or later a client or your finance director will ask what AI saved, and without that note there is nothing to compare against.
New survey figures suggest many marketing teams are already in that position. Nearly half have moved work to AI in the past year, and around two in three either don't measure the return or don't know whether anyone does.
What the survey found
Basis, a company that sells advertising technology including AI tools, surveyed 184 marketing and advertising professionals at agencies and brands for its fourth annual AI and the Future of Marketing report, released with a press release on 19 October. It is a small sample run by a vendor, and Basis's summary and press release give no sampling method or margin of error, so read the figures as a direction of travel.
47.8% said their organisation replaced human tasks with AI in the past 12 months, up from 27.9% a year earlier. The survey asked about tasks, not jobs: handing over a task doesn't mean someone lost their role.
On the return, 65.1% said they don't measure AI's return at all, or don't know whether anyone at their organisation does. Only 6.5% said they evaluate their AI investments against defined metrics and benchmarks.
Choosing well and measuring well are the same job
Our position is that the right first AI project is the most painful, most repeated piece of work in the team, picked by what it costs and how often it happens, rather than by the most impressive demo. This survey adds a reason to take the baseline seriously.
If you choose by cost and frequency, you have to find out what a task costs and how often it comes round. That is your baseline. You write it down in the act of choosing, and the case for the return follows from it. Teams that pick by what looks exciting skip the step, because nothing in their decision required it.
In our view, teams that automate first and plan to measure later will find it hard to prove anything. Once the work has changed, the old version is gone, and "it used to take ages" won't hold up in a budget meeting.
Lloyds Banking Group is one example of testing AI against a human baseline. As The State of Brand reports, Lloyds tested AI on a real campaign brief in a project it called Turing: three ring-fenced teams worked on the same brief, one without AI, one helped by AI and one using AI alone. Separately, Lloyds says generative AI delivered £50m of value across the group in 2025 and expects more than £100m this year. Those are Lloyds' own group-wide figures, and the method behind them isn't shown.
Lloyds' marketing chief, Suresh Balaji, has also warned that some of the bank's AI content experiments risked degrading the brand. In our view, that is the other half of a baseline's value: beyond justifying the spend, it shows which work to hand over and which to keep.
What it means for agencies and in-house teams
For agencies, clients who see AI in their retainers will ask what it saved them. An agency that recorded hours and turnaround before it changed a task can answer in the client's terms. One that didn't is left defending its fee with anecdotes.
For in-house teams, AI spending that can't show a return is the easiest line for a finance director to question. A short baseline on the few tasks you've handed over turns "we think it helps" into a figure the budget meeting can use.
There is a fair objection here. Measuring everything slows adoption down, and nobody wants a timesheet exercise before every experiment. So don't measure everything. Measure the handful of tasks that matter most, the most repeated and most costly ones, which are the ones you should be starting with anyway.
What to do this week
- Write down the cost of your three most repeated tasks before you change them. Hours per week, cost per piece, time from brief to delivery. Ask the people who do the work, not the project plan.
- Agree what "saved" means. With the client or your finance lead, pick one measure per task: hours, turnaround or cost per output. One is enough.
- Put a monthly check in the diary. Same tasks, same measures, compared with the baseline you wrote down.
- Report savings in the client's terms. Faster launches and a lower cost per asset mean something in a client meeting. The number of AI tools rolled out doesn't.
Which task would go first on your list of three, and how would you know if handing it to AI had paid off?
Jules, founder of WYNStrategy Director helping agencies and marketing teams use AI properly. More about Jules