AI in Campaign Management: The Practical Guide.

Many opportunities already exist, but they are not being taken advantage of. Before thinking about the future, it’s worth taking an honest look at the status quo.

What you should check:

• Do you use automated bidding, or do you still manage your bids manually?
• Are budgets automatically allocated across channels and target audiences?
• Do you use Dynamic Creative Optimization?
• Does your reporting setup automatically detect anomalies?

Optimization is only as good as the metric used to measure it.

Here’s what you can do specifically:

• Question the attribution model you’re using: Which touchpoints are being under- or overvalued as a result?
• Check to see if your KPIs reflect not only performance but also brand impact
• Test incrementality regularly using geo-experiments and lift studies
• Implement marketing mix modeling to understand the overall contribution of your channels

The number of data points is decreasing, and platform ecosystems are becoming more closed. Those who don’t rely on their own data now are strategically relinquishing control.

Here’s what you can do specifically:

• Get an overview of the first-party data you already have
• Identify touchpoints that are suitable for data enrichment
• Implement server-side tracking and CAPIswherever possible
• Proactively ensure compliance with data protection regulations; the GDPR is no reason to wait and see
• Enable first-party data through Customer Match and similar mechanisms

Programmatic advertising involves various intermediaries between advertisers and publishers, each offering its own added value. Those who actively manage their supply path can buy more efficiently.

Here’s what you can do specifically:

• Analyze which SSPs and purchase paths you’re currently using
• Minimize redundant paths and opt for direct deals where appropriate
• Check inventory quality regularly
• Ask your agency: How much of my budget goes toward working media?

Reach does not equal relevance. A technically flawless ad delivery is not the same as a contextually appropriate placement.

Here’s what you can do specifically:

• Don’t just define what you’re ruling out, but also where you want to be actively visible
• Establish a regular monitoring process
• Make sure someone actively decides whether the placement is a good fit for the brand

In the past, there was one theme and one format per campaign. Today, hundreds of variations are needed at the same time—but not without a strategy.

Here’s what you can do specifically:

• Define a clear hierarchy of messages: What is the core message, and what are the variations?
• Keep your target audience setups separate; otherwise, the algorithm will learn the wrong things.
• Schedule ongoing quality checks: Who reviews the top options on a regular basis?
• Specify when a creative should be paused, even if it’s still performing well, but, for example, no longer aligns with the brand.

There are some analyses that used to be simply too time-consuming. Not because no one wanted them, but because they would have taken hours. That’s exactly what’s changing right now.

Here’s what you can do specifically:

• Set up automated alerts
• Analyze regularly at the segment level: Which smaller target groups are performing above average?
• Keep your exclusion list up to date: Which URLs and apps are producing poor results?
• Make sure that someone not only looks at the data, but also interprets it and makes decisions

How much of the budget should go to each channel? In many companies, this question is still answered the same way it was ten years ago: based on past experience, gut feeling, and last year’s budget plus or minus x percent.

Here’s what you can do specifically:

• Simulate scenarios: What happens if you shift the budget from Channel A to Channel B?
• Use historical data to identify seasonal patterns
• Plan budgets flexibly, but with clear rules for when and how they should be adjusted
• Check your channel mix regularly: Is it based on current insights or on habit?

The last touchpoint isn’t the whole story: The reports show conversions. But would that person have converted even without seeing the ad? Without asking this question, you’re only optimizing for numbers, not for impact.

Here’s what you can do specifically:

• Set up geo-experiments: The campaign is running in Region A, but not in Region B; the difference shows the actual impact
• Tests Using Lift Studies: The test group sees the ad, while the control group does not
• Accept uncomfortable results: If incrementality shows that a campaign was less effective than expected, that’s not a setback—it’s valuable information.

AI automates many tasks. But transparency doesn’t happen on its own.

What you can ask for:

• What objective is the algorithm currently optimizing for, and is that still the right one?
• In which areas does your agency actively intervene, and in which areas does AI make the decisions?
• How can we ensure that brand safety is more than just a checkbox in the tool?
• When was the last time budgets were actively reallocated because performance required it?
• How does the agency explain to me exactly what AI does in my campaigns?

AI decides in real time who sees an ad, how much it costs, when it’s served, and which ad format is the best fit. From marketing leads to the CEO, everyone involved in campaign results should understand the basics. The better you understand them, the more precisely you can manage your campaigns.

Here’s what you can do specifically:

• Explain internally exactly what decisions AI makes in your campaigns
• Clear Responsibilities: What does AI decide, what does the agency decide, and what does your company decide?
• Create opportunities for collaborative learning: AI is evolving rapidly, and the best results come from testing and sharing ideas

For us, AI isn’t just an add-on. It’s an integral part of how we plan, manage, and optimize campaigns.

What this means in concrete terms:

• We use AI for media planning, bidding, and budget management; we define the strategy behind it together with our clients
• We rely on automated reporting and analysis; we provide the interpretation and recommendations
• We test new approaches together with our customers because AI is evolving rapidly and learning is a collaborative process

A concrete example is edify, our framework for agentic performance marketing. Put in a briefing, get hundreds of on-brand assets in return. People provide the creative direction; edify handles the scaling.

We use AI where it adds real value. We use the flexibility this creates to focus on what makes for good consulting: strategy, context, and a genuine understanding of our clients’ goals.