For many years, the work of a marketing specialist was a mix of analysis, experience, intuition and a lot of manual work.
Open the platform. Check the spend. Look at the conversions. Compare yesterday with last week. Try to understand why a campaign is spending too much, why the CPA is going up, why impressions suddenly doubled, why a conversion stopped tracking, why Google is showing ads for something completely irrelevant.
Then open another platform and do the same thing again.
The problem is not that marketing people are lazy. The problem is that most of their energy is consumed by collecting information instead of using it.
This is the part where AI can be useful. Not because it should take decisions by itself. And not because we need another chatbot that tells us generic things about “optimising the funnel”. But because it can do the homework.
Recently I started working on a small system connecting Google Ads data to a dedicated Google Cloud environment. The system can read campaign data every day: spend, budget, impressions, clicks, conversions, conversion tracking setup, search terms, asset performance, bidding targets and changes made in the account.
Then it creates a structured view of what is happening.
For example, it can notice that a campaign had a sudden increase in impressions but not in clicks, that the CPA increased after a target CPA or budget change, or that a Performance Max campaign is receiving traffic from irrelevant search themes. It can also look at which creative assets are generating conversions at an acceptable cost and which ones have spent enough money without producing enough.
This does not mean that the system should go and change the campaign automatically. I am not a big fan of giving full control to a machine just because it can press buttons faster than a human.
A campaign is not a spreadsheet. There are commercial priorities, seasonal effects, brand considerations, client relationships and information that Google will never have. Sometimes a campaign with a bad CPA should stay live. Sometimes a campaign with a good CPA should be stopped. The numbers are important, but they are not the whole story.
The useful model is different.
The machine reads the data, applies a set of rules and creates a proposal. The proposal explains what happened, why it could matter, how urgent it is, what the possible impact could be and which Google documentation or account data supports the suggestion.
Then a marketing specialist reviews it.
The specialist can approve it, reject it or modify it. Most importantly, they can explain why. The decision is documented and becomes part of the history of the account.
In a more mature version, every meaningful suggestion can become a Jira ticket assigned to the person responsible for that campaign. This is not about monitoring people in a paranoid way. It is about making the work visible, reducing unnecessary subjectivity and creating a consistent process.
Instead of asking “did anyone check the campaigns today?”, you can see what the system found, what the specialist decided and what happened afterwards.
The same logic applies to reporting.
Reporting is often a painful ritual. Someone exports data, someone updates a presentation, someone asks why a number changed, and after a few days the discussion is already outdated. A proper system can collect the data daily, store it in a central place and produce recurring reporting around the questions that actually matter.
What changed? Why did it change? What should we investigate? What should we leave alone?
The last question is probably the most important one. Not every change requires an action. Especially with automated bidding, continuous manual intervention can easily make performance worse. Sometimes the best recommendation is simply: do not touch this campaign for the next few days and let the algorithm stabilise.
This is how I see the practical use of AI in marketing today.
Not a replacement for specialists. Not an oracle. Not a PowerPoint with the word “agent” on every slide.
Just a system that does the repetitive work, brings the relevant evidence to the table and leaves human beings with more time to think, challenge assumptions and make better decisions.