AI’s ability to achieve results for marketers is no longer in question, and hasn’t been for some time. At this point, most online retail businesses would be non-functional without this technology: it is baked into the business, with everything from copy-generation to budget allocation bound up with AI decision-making and automation.
Most of the time, this is an all-upside situation. It is the rare online business that has suffered at the hands of AI technology, except perhaps for those selling tools that AI directly competes with. And yet, as many have come to learn, there is genuine peril in surrendering full control to the machines.
Let’s say you notice a certain automated campaign is underperforming by a few percentage points compared to previous months. You dig into the data and notice the AI has temporarily switched your target demo, raising the age range or focusing more attention on a specific region. You want to reverse course, but you also don’t have enough information to know what the AI was thinking.
Why the change? What does it mean?
In fact, the opposite situation can be just as troubling. You notice the AI has switched the target demographic for your campaign, with great success. Clearly, they’ve tapped into market insights that might benefit other sectors of your business. Once again, though, you’re in the dark—you don’t know why they’ve made a certain decision, and so you can’t put this new information to use.
It is for this reason that recent discussions around AI in marketing have shifted. Previously, the imperative was more AI—finding new ways to incorporate the technology into your systems. Today, the conversation is about finding ways to better understand AI’s decision-making. Hence the rise of what is known as explainable AI.
Why explainable AI matters
Traditional AI systems operate like a black box. You input your campaign parameters, and they get to work optimizing your settings: finding the right words to land the right demographic at just the right price point. However effective their results may be, they remain mysterious to the marketer because you can’t see the work.
That matters. It matters when campaigns are going well, and it especially matters when campaigns are going poorly. Let’s say the AI has suggested you target men between the ages of 42 and 48 interested in higher-end menswear living in coastal cities. Let’s say this age range is a bit higher than your typical demographic. That might be okay, especially if it’s bringing in sales—but you need to understand why the AI is making this decision before you can fully sign off on it.
For instance: is there solid past conversion data to back this decision up? Have your past campaigns—despite your initial intentions—somehow appealed more to this demographic? These things are useful to know. And they are essential to know if the campaign fails. Because if and when higher-ups ask why ROI has fallen, you cannot simply say: the computer did it. The final decision-making, ultimately, rests with the marketer.
How explainable AI works
Explainable AI refers to AI systems that can provide, in plain language, the rationale behind each individual decision that is made. For an advertising or marketing professional, that could mean explaining, step by step, why the system chose urgent messaging in a particular campaign. For example, it may be the case that urgent appeals performed better 42% of the time in past campaigns with similar parameters. It could also explain why it targeted middle-aged coastal menswear shoppers in the first place — because patterns from previous campaigns show that this demographic is significantly more likely to convert on this particular category of product.
What explainable AI is offering here is the full range of factors that went into each decision—the reasoning (why it did what it did), the context (to what extent did circumstances justify the change?), the relevant evidence (what’s the data looking like on each decision?).
Now you’re learning. Now you’re actually armed with information you can use. Which means you can override the system, depending on your brand’s top-level goals. Or finesse your other campaigns in line with this new information, if this particular AI-facilitated campaign is producing quality results.
Among the million daily micro-transactions between a brand and its customers, rich stores of data are constantly being generated. AI is good at channeling those flows into high-return outcomes, but it is not infallible, and at the end of the day, it is the marketer that is responsible for every penny of spend.
Transparency as a buying criterion
Previously, marketers were in the market for the AI equivalent of muscle cars—AI tools that could go fast, that could scoop up the highest returns possible with the greatest degree of automation. That is set to change, as explainable AI increasingly becomes the reigning paradigm for the industry. How the AI actually works is, more and more, likely to factor into purchasing decisions. In consequence, one can easily imagine a scenario in which the next ad-tech battleground is over who can offer the “most” AI, but rather who can offer AI that is authentically explainable.
The human-plus-AI model is what will win here. In other words: AI is a kind of hyper-competent colleague. If you had a human colleague who did great work 90% of the time and abysmal work 10% of the time, and who furthermore refused to explain themselves when their projects refused to pan out, you probably wouldn’t think very highly of them. The same, going forward, will apply to ad-tech AI tools. Black-box automation is, increasingly, a thing of the past. Explainable automation is the way forward.
By Mitsunaga Kikuchi, CEO and Founder, Shirofune
Mitsunaga Kikuchi is the founder and CEO of Shirofune, the digital advertising management tool that automates ad campaigns through an easy-to-use interface for management, budgeting, monitorig and analytics. Mitsunaga recently won the 2025 ANA MarTech Gold Award for individual marketing solutions provider and the 2025 AdWeek TechStack Award.
