AI is no longer the new kid on the adtech block. It is baked into the way we buy media, measure attention, and personalise experiences. Yet the more embedded AI becomes, the more important it is to handle it with care.
One of my favourite analogies lies in archery: if your aim is a single degree off at the moment the arrow leaves the bow, it will veer wildly off target at inhuman speed and possibly miss the target altogether. The same is true of AI: a tiny calibration error in the model or the data can snowball into reputational damage, wasted budget or even regulatory pain.
The promise and the peril
The promise of AI is, of course, huge. It can already surface audiences we would never have found manually, optimise thousands of creatives in real time, and strip out redundant impressions. In the next decade, it will touch every corner of our lives - from avatar GPs to real-time trading desks that learn on the fly.
The peril is equally real. We all remember search engines that summarised content so well that publishers lost the click-throughs altogether, or product feeds that paired a "kid’s birthday gift" query with the sort of results only a horror fan would love. My own email assistant decided my messages were "too nice", and it didn’t work out. These are not just embarrassing glitches. When AI scales, mistakes scale with it.
Human equity: your competitive edge
If every platform gains access to comparable models, where will differentiation come from? Not from brute processing power, but from human equity - the curiosity, empathy, and commercial judgement that steer the algorithms. Your culture, your values, your secret sauce and the questions your team chooses to ask are the inputs no rival can copy. This is where the intersection between AI and human equity really delivers traction and results.
In practice that means:
- Define the outcome first. What problem are you trying to solve and for whom? Shiny tools are seductive, but unless they map to a clear customer need, they will become expensive toys. Efficiencies are always welcome but not at the cost of human empathy and nuance
- Interrogate the data. Is it current, complete, and free from hidden bias? Models trained on skewed data will amplify that skew.
- Keep the human(s) in the loop. Set guardrails, audit regularly, and give people the authority to pause the machine when something looks off.
- Educate everyone. From sales to finance, your teams need a working knowledge of what the AI is doing so they can spot oddities early.
We’ve already seen plenty of practical examples of humans and AI singing in tune. As my fellow panelist at MAD//Fest, Samsung Ads’ Lauren Barnett, highlighted, a recent CTV campaign that paired first-party viewing data with an ML look-alike model increased app opens by nearly 60%, while cutting cost per app open by 13%.
However, the clever bit was not the algorithm itself, but the planners who fed it fresh audience signals every 24 hours, and rewrote the brief when the data suggested a previously hidden sports-mad segment. Without that creative nudge, the model would have stayed in its lane and delivered average results.
Six rules for responsible AI adoption
Based on our experiences so far, the six steps to success with an AI-assisted campaign would be:
- Start small, scale fast. Pilot your campaign in a contained channel, measure hard outcomes, then widen the net.
- Build multi-disciplinary squads. Pair data analysts with strategists, creatives, and compliance experts. Diversity of thought reduces model myopia.
- Stress-test the edge cases. Ask what happens when a parameter drifts, a feed breaks or a brand-safety list lags. Run tabletop exercises before the crisis hits.
- Budget for monitoring. Set aside a small proportion of the project cost for ongoing auditing and fine-tuning. If you skimp here and the archery analogy comes back to haunt you.
- Champion transparency. Tell customers when AI is involved, why it benefits them, and what safeguards exist. Trust is the currency that buys you permission to innovate.
- Work with Partners like Limelight Inc who can help you to deliver the values, vision and results that best represent your specific narrative as a business, don't settle for ubiquitous technology devoid of human direction
Looking ahead
We are at the very beginning of the AI revolution. The winners will be those who embrace the technology yet keep a firm hand on the tiller. In ten years, we will look back at today’s tools the way we now look at dial-up modems: quaint but groundbreaking for their time. What will endure is the insight that machines, however powerful, serve human goals. Our job is to articulate those goals with precision and integrity.
As marketers, let us harness the tech, drive the tech, and retain the human touch. Get that balance right and AI becomes an accelerator rather than a risk factor. Get it wrong and the arrow misses the target - sometimes spectacularly.
Copyright 2025. Featured post by James Macdonald, Limelight Inc.