Digital advertising is no longer just media bought through a screen. Increasingly, it is a software system that decides who sees an advert, how much is paid for the impression, which version of the creative appears and whether the campaign should spend more or less at that moment. IAB UK’s 2025 Digital Adspend research found that 48% of industry respondents saw AI as one of the most important forces shaping advertising over the next decade. That figure helps explain why agencies and advertisers are paying so much attention to automation. The attraction is obvious: software can process more data, test more variations and react faster than a human team working manually. The risk is equally obvious: the more decisions are delegated to systems that advertisers do not fully understand, the easier it becomes to scale the wrong objective, depend on opaque platforms and mistake machine confidence for business insight.
Why automation changes the economics of media buying
Traditional media buying often involves selecting a channel, negotiating inventory, agreeing a schedule and committing budget in advance. Digital platforms work differently. Search, social, retail media and programmatic systems can auction impressions in real time, adjust bids continuously and move spend according to performance signals.
This changes the economics because the cost of making many small decisions falls dramatically. A human buyer cannot evaluate millions of individual impressions one by one. Software can. It can compare device, geography, audience segment, time of day, previous behaviour and conversion probability before deciding whether an impression is worth buying.
The advantage is scale. A relatively small team can manage campaigns across thousands of keywords, audience groups and creative combinations. Automation also reduces repetitive manual work, which allows specialists to focus more time on strategy, creative direction and commercial interpretation.
The disadvantage is dependency. Once a campaign relies heavily on automated bidding and platform optimisation, the advertiser becomes dependent on the rules and data of that platform. Changes to auction mechanics, targeting options or reporting can affect performance quickly. The advertiser may know what outcome the system produced without knowing exactly why it produced it.
That trade-off is central to modern digital marketing. Software makes media buying more efficient, but efficiency comes with less direct control over many individual decisions.
How AI expands testing, targeting and creative production
AI is widening automation beyond bidding. Generative systems can now help produce headlines, image variations, video concepts, landing-page copy and audience ideas. Campaign teams can create many more versions of an advert than would have been practical with a fully manual process.
The benefit is faster experimentation. An advertiser can test different messages for different audience segments and learn which combinations appear to work. This can be especially useful for ecommerce, where product ranges are large and campaigns need frequent creative refreshes.
AI can also help analyse campaign data, identify anomalies and suggest budget reallocations. Used carefully, this can reduce the time required to find obvious underperformance. It can support smaller teams that do not have large in-house analytics departments.
The drawback is that faster production can create more mediocre content rather than better content. If every advertiser uses similar tools trained on similar patterns, creative can become generic. Personalisation may increase in quantity while distinctiveness falls.
There are also quality and brand-safety risks. AI-generated material can contain errors, unsuitable claims or imagery that does not fit the brand. Human review remains essential. Automation lowers the cost of producing a mistake just as effectively as it lowers the cost of producing a useful variation.
This is why AI should be viewed as a production and decision-support layer rather than a substitute for strategy. It can create options quickly, but the business still needs to decide what it wants to say, who it wants to reach and what outcome genuinely matters.
The black-box problem and why bad inputs can scale quickly
Automated platforms optimise towards the signals they receive. If those signals are good, automation can be powerful. If they are poor, the system can pursue the wrong objective with impressive efficiency.
Consider lead generation. A company may tell a platform to maximise completed forms. The algorithm may learn how to find people who submit forms cheaply, but those leads might have low purchase intent. Cost per lead falls and the dashboard looks healthier, while sales quality deteriorates.
The same issue appears with ecommerce conversion tracking. If purchases are misconfigured, duplicated or attributed incorrectly, automated bidding can learn from distorted data. The system is not being irrational; it is doing exactly what it was instructed to do with the information provided.
The advantage of automation is consistency and speed. The disadvantage is that errors propagate quickly. A human buyer making a poor decision might waste part of a budget. An automated system can repeat that poor decision thousands of times before anyone notices.
Opacity makes this harder. Platforms increasingly use models that combine many signals, and advertisers may not receive a complete explanation of why specific users or placements were chosen. This can make troubleshooting difficult, particularly when performance changes suddenly.
The practical response is stronger governance: accurate conversion tracking, clear definitions of valuable outcomes, routine audits, incrementality testing and human review of major automated changes. Automation should reduce manual workload, not reduce accountability.
Where traditional media still offers useful simplicity
Traditional media is often criticised for being less sophisticated, but simplicity can be an advantage. A radio campaign, print placement or outdoor booking may involve fewer software dependencies and fewer hidden decision layers. The advertiser usually understands what was bought, where it appeared and for roughly how long.
That does not make traditional media easier to measure. Audience response is usually less granular, and optimisation is slower. Once a print run is complete or a billboard is booked, the campaign cannot be changed as rapidly as a digital ad set.
But the simpler buying model can make governance clearer. There is less risk that a hidden algorithm silently changed audience composition or bid strategy overnight. For brands in highly regulated categories, that predictability can be valuable.
Traditional channels can also provide context that automated digital systems struggle to guarantee consistently. A full-page placement in a trusted publication or a prominent outdoor site has a known environment. Programmatic digital inventory can vary much more widely, which creates brand-safety and quality-control challenges.
The disadvantage is flexibility. Traditional media offers fewer opportunities for rapid testing, personalised creative and individual-level targeting. It can also be expensive to experiment at scale.
This means the comparison is not simply advanced software versus outdated media. It is a trade-off between automation and transparency, flexibility and predictability, granular targeting and broader contextual reach.
How agencies should use AI without surrendering judgement
The strongest use of AI and automation is not to remove people from the process but to move people to the parts of the process where judgement matters most. Software is excellent at repetitive optimisation, pattern recognition and generating variants. Humans remain responsible for business objectives, brand meaning, ethical boundaries and deciding whether the reported outcome is commercially valuable.
For agencies, this means being able to explain not just which tools are being used, but what those tools are optimising towards. Clients should know which conversion events matter, what attribution model is being used, which parts of the campaign are automated and where manual controls still exist.
The benefits are considerable. Faster testing can improve learning. Automated bidding can reduce wasted labour. AI-assisted creative can make campaigns more responsive. First-party data can help personalise messages more effectively.
The disadvantages are equally real. Martech stacks are expensive to integrate, platform skills need constant updating, privacy rules constrain data use, and black-box systems can make performance harder to explain. Agencies that promise effortless AI-driven growth risk replacing one kind of marketing jargon with another.
The sensible model is supervised automation. Let machines handle volume, but keep humans responsible for direction. That means checking data quality, testing whether conversions are incremental, reviewing creative, monitoring brand safety and challenging recommendations that do not make commercial sense.
AI is reshaping digital advertising because it reduces the cost of making and testing decisions at scale. That is why budgets continue to follow software-driven media. But automation is not automatically intelligence, and optimisation is not the same as strategy. The firms that benefit most will be those that use the technology aggressively enough to gain speed, but cautiously enough to notice when the machine is confidently optimising the wrong thing.
Source: IAB UK Digital Adspend 2025 and associated AI/automation commentary.