In this guest article, Jessica Michen, Co-Founder & COO at Grasp, a media governance and data quality tool owned by programmatic advertising platform MiQ, outlines the consequences advertisers face if they’re using ‘bad data’ and the role of ‘preventative governance’ in helping solve it.
Agentic AI is quickly moving from industry talking point to operational reality. In media buying, the appeal is obvious: faster planning, faster optimisation, faster execution, and systems that can act with less human intervention.
For an industry built on speed, that sounds like progress. But there is a less comfortable question we need to ask before we hand more decisions to machines: what happens when the data they are acting on is wrong?
Because in media, the problem is rarely a lack of automation. The problem is that too much of the industry’s automation is being built on shaky foundations.
We have all heard the promise. AI agents will monitor campaigns, adjust budgets, recommend audiences, shift spend, and identify underperformance in real time. In theory, this should make media buying more efficient and more accountable. But AI systems are only as good as the inputs they are given. If campaign data is mislabelled, incomplete or inconsistent at the point of setup, then automation does not solve the problem. It makes the problem move faster.
That is the risk now facing advertisers. We are entering an age where a small operational error is no longer just a reporting headache. It can become the basis for automated decisions made across markets, platforms and budgets.
The first domino
Taxonomy is not the most glamorous part of advertising. It is not the creative idea, the media strategy or the shiny new AI tool. But it is the structure that allows campaigns, ads and data to be named, organised and understood consistently.
Get it right, and teams can see what is working across platforms and markets. Get it wrong, and performance data starts to fracture. A misplaced space, an incomplete field or a slightly different naming convention can break the link between what was planned, what was bought and what was measured.
In a manual world, that was already a serious issue. In an AI-assisted world, it becomes much bigger.
If an AI optimisation system is being fed incomplete or inconsistent campaign data, then its recommendations will be shaped by that flawed picture. It may shift the budget away from the wrong activity, double down on a misleading signal or fail to identify what is actually driving performance. The machine may be acting quickly, but it is not necessarily acting intelligently.
This is why taxonomy should be treated as the first domino in a clean data strategy. If it falls, everything that follows becomes less reliable.
The World Cup test
The 2026 FIFA World Cup will be one of the clearest tests of this new reality.
Global sporting events create exactly the kind of environment where AI-enabled media buying will be expected to prove its value. Brands will be running complex campaigns across dozens of markets, platforms, agencies and formats, often with very limited windows to optimise performance while the event is live.
The scale is enormous. Grasp analysis suggests that up to $3.9 billion of World Cup-related advertising activity could be at risk of running with compromised campaign data due to inconsistent naming and poor taxonomy practices.
That does not mean the money simply disappears. It means brands may not be able to see clearly how that money is performing. They may struggle to compare markets, assess channels, optimise in real time or understand which campaign elements actually drove outcomes.
During a long brand campaign, there may be time to spot the issue and correct it. During a World Cup activation, there often is not. By the time teams realise the reporting is broken, the moment may already have passed.
This is the problem with treating data governance as a clean-up job. Once the campaign is live, the budget is moving, the platforms are learning, and increasingly, AI is making or shaping decisions. Trying to correct the foundations after that point is incredibly difficult.
The new rules of AI-era media buying
As AI agents take on more responsibility in campaign execution, brands and agencies will need to be much clearer about where judgment and accountability sit. If an automated system makes a poor optimisation decision because a campaign was labelled incorrectly, the answer cannot simply be “the machine did it”.
Agentic AI may change how media decisions are made, but it does not remove the need for experienced oversight. It raises the stakes on it. The faster a system can act, the more important it becomes to have people who understand the campaign, the market and the context behind the data.
That is where preventative governance has a critical role to play. Its value is not in taking humans out of the process, but in giving them stronger foundations to work from: cleaner inputs, clearer alerts and more reliable data from the start. When campaign naming, taxonomy and set-up checks are embedded directly into the workflow, media teams are not left trying to catch every possible error manually. They are better equipped to read the signals that matter.
Because good media professionals do far more than press buttons. They know when a spike in performance might be a tracking issue, when a market is behaving unusually because of local context, or when a campaign is technically “optimised” but commercially misaligned. AI-enabled governance strengthens that judgement by making sure the signals they are reading are accurate, consistent and trustworthy.
That becomes especially important during moments like the World Cup, when campaigns are moving quickly across multiple markets, platforms and agencies. In that environment, speed without trust is not progress. It is just risk at scale.
The future of media buying will almost certainly be more automated. But the brands that benefit most will not be the ones that remove people from the process. They will be the ones that use AI to empower them, giving teams the safeguards, data and confidence they need to make better decisions at speed. Ultimately, the goal should not be media buying without humans. It should be media buying where human judgment is better supported.
