AI is becoming a much bigger part of not only our everyday life, but also Amazon advertising, and that is a good thing in many different aspects. It can analyze large amounts of data quickly, identify trends, help with bid and budget decisions, and give advertisers another layer of insight when managing complex accounts, but AI is only as useful as the goals and context you give it.
That is where I think brands need to be careful, especially when return on ad spend (ROAS) becomes the main or only key performance indicator (KPI).
ROAS Is Important, but I Don’t Believe It’s THE GOAL
I look at ROAS constantly when managing Amazon accounts. It matters, but it is rarely the only thing a client cares about. A brand may be focused on profitability, total sales growth, inventory levels, a new product launch, market share, organic ranking, moving a specific product line, or a number of different other equally important metrics. Often, several of those priorities are happening at the same time.
In fact, every time I begin working with a new client, the first meeting we have together, we spend the majority of the call going over short- and long-term goals. I initially let them tell me what their goals are, and almost every single time, the client gives a ROAS (or any return metric they focus on) goal as either their top goal or their only goal. As we continue the conversation and relationship, many times those goals are updated due to inventory, market share, competition, and organic ranking, etc.
That context changes what a “good” advertising decision looks like.
A lower ROAS campaign may be supporting a new product launch, high-volume general keywords, competitors’ names or ASINs, or products with better margins than others in your catalog, etc.
Looking only at ROAS can miss all of that.
AI Learns What You Tell It Matters
The more we use AI, the more important this becomes.
If every conversation with AI is about improving ROAS, lowering ACOS, cutting inefficient spend, and finding the campaigns that look worst on paper, we are repeatedly telling it that efficiency is what matters most. Over time, that shapes the recommendations we get back. In addition, I can promise you everyone using AI for paid advertising management has asked their AI to help lower their ROAS and increase revenue.
AI may recommend reducing bids on expensive keywords, cutting non-branded campaigns, or moving more money toward branded traffic because those actions can improve ROAS. From a purely ROAS-driven perspective, those recommendations might make sense. From a business perspective, they probably do not.
The problem is not necessarily that AI made a bad recommendation, but rather that it doesn’t clearly understand the entire picture.
Test AI on Something You Already Know
One thing I often tell people who assume an AI answer must be correct is to use it on a subject they are already an expert in.
Then push it. Ask it more in-depth questions.
You will quickly see that AI is incredibly impressive but far from perfect.
It knows a lot about baseball and the stats around the game, but it doesn’t understand the little things about the game that you develop over years of playing. It doesn’t understand things like clubhouse chemistry, veteran leadership, and some players being able to be calm under pressure while others can’t.
AI knows a tremendous amount at the surface level. The deeper you go, the more likely you are to notice missing context, oversimplified recommendations, or answers that technically make sense but are not what an experienced person would actually do.
I see the same thing with Amazon advertising because I work in these accounts every day. I can usually tell when an AI recommendation is helpful, when it needs more information, and when I would ignore it completely.
Someone with less advertising experience may not know the difference. That does not make AI less useful; it just means experience is still important, especially when click costs continue to rise, and the competition without a doubt will be leveraging AI and telling it to focus on ROAS and growth.
The Human Piece Is Context
As an account manager, I am not only looking at campaign data. I am also thinking about what the client has told me matters to their business.
Maybe profitability is the priority this quarter. Maybe they are willing to sacrifice efficiency to grow a category. Maybe inventory is becoming an issue. Maybe sales are down year over year, and advertising is only one piece of the problem. That information is not sitting in some report and is continually changing.
AI can help analyze all of it, but someone still needs to know what information matters, what questions to ask, and when a recommendation does not fit the larger business goal.
AI Works Best With Experience Behind It
I do not see AI replacing experienced account management. I see it making experienced account managers better. I leverage AI to do my job but only in certain areas. Pulling information that would take me hours or days to compile can now be done in seconds or minutes. But speed and access to information are not the same thing as experience.
The strongest approach is combining both.
Use AI to help analyze the data and surface opportunities. Then have the experienced people provide the context, define the real goals, and challenge the recommendations when necessary.
Because the goal should not be teaching AI how to get the highest ROAS possible, it should be teaching AI what success actually looks like for the business.
