AI in programmatic: progress or just a faster way to make the same mistakes?

James MacDonald, Co-Founder and CRO at Limelight says companies should think carefully before rushing into AI deployments

AI has dominated the news for all the wrong reasons lately, with dark predictions of an AI-fuelled apocalypse coming from people with enough credibility and experience to demand that we take them seriously.

In the day-to-day programmatic world, however, the outlook is far more pragmatic. Publishers and advertisers are putting AI to work to streamline workflows, surface clearer performance signals, and gain greater visibility into what’s working and what isn’t. The technology is already influencing everyday decisions around bidding, supply paths and inventory quality.

Even here, though, caution is warranted. In the rush to embrace AI, some people are losing sight of what it is and what it isn’t; of what it can do, and what it can’t.

AI is being introduced across programmatic with the promise of better outcomes and smarter trading. And without a doubt, when AI is successfully integrated into a company’s advertising process, it can deliver greater efficiency, improved performance and easier scaling on both the buy-side and sell-side.

Also Read: The Marketer’s Questions Haven’t Changed. The Industry Has.

But AI can achieve nothing on its own. It is a powerful tool, not a strategy. It creates real value only when built on quality data, clear objectives and human oversight. It will not fix broken incentives or inefficient supply chains by itself.

If the underlying data is flawed or incomplete, AI-driven automation simply scales the problem. Optimise against the wrong KPI and the system will chase it with ruthless efficiency. Feed it noisy or biased inventory signals and it will amplify low-quality supply. Without tighter control and clearer signals, AI risks accelerating inefficiency, rather than solving it.

In other words, the aim should be to support human-directed strategy and goals, rather than attempting to replace them.

The three factors that decide whether AI works

Here’s the advice I would give to any company looking to deploy AI successfully in its advertising operations.

• Start with the problem, not the technology. The surest way to waste an AI investment is to begin with the tool and then hunt for problems it might solve. Reverse the order. Identify the real issues and root causes in your current processes, along with the objectives you are struggling to achieve. Only then should you ask whether AI can help - and stress-test the idea thoroughly before committing.
• Get your data in order first. You cannot properly diagnose problems or set meaningful objectives without reliable insight into your systems, processes and workflows. Data is the foundation. An AI deployment built on flawed or incomplete data is a recipe for failure.
• Let AI advise, let humans decide. No matter how sophisticated the platform, AI cannot grasp editorial values, audience nuance or the strategic intent behind a publisher’s brand or an advertiser’s campaign. Human oversight remains essential - to set direction, contextualise outputs and uphold standards that pure optimisation will never prioritise on its own.

In practice, the difference usually comes down to a few unglamorous habits. The teams with a coherent AI strategy typically treat data quality as an ongoing operational priority rather than a one-off project. They judge success by more than short-term efficiency, taking brand safety, longer-term yield and audience experience into account. And they make sure people stay in the loop, so the technology remains accountable to strategy rather than quietly rewriting it.

The bit the technology can’t do

At Limelight, we help companies navigate the programmatic landscape by combining AI with experienced human support. Our Adaptive Rules Centre leans heavily on AI to enable our clients to customise and optimise ad performance through automation. Users can tweak a variety of parameters, including the number of queries per second (QPS), geotargeting, fill rates and bid rates. This allows them to scale demand for supply sources that are performing well, while throttling demand for those that aren’t.

Even with this level of automation, real human expertise remains only a phone call, an email or a WhatsApp away. Clients often tell us that the ability to speak to someone who has seen hundreds of campaigns across different market conditions is still the difference between a tool they use and a platform they trust.

AI is already delivering genuine efficiencies and performance gains for many in programmatic. But to extract the greatest value, treat it as what it is: a powerful tool that helps people make better, more informed decisions - decisions guided by AI, but made by humans.