Most Amazon Ads automation tools position AI as the destination. Hand the algorithm the keys, let it optimize, report on outcomes. The pitch is appealing: less work, better results, decisions made at machine speed. We heard it. We considered it. And then we built something different.
Off Hours is built on rules you define, not algorithms you trust. That was a deliberate choice, and it shapes everything about how the product works. A dayparting rule says "pause these campaigns from midnight to 6am on weekdays." A performance rule says "if yesterday's ACoS is above 35%, send me an alert and do nothing else." You wrote those rules. You can explain them to a client. You can change them tomorrow morning before the engine runs. The decision is always yours.
The auditability problem
When something changes in your account, you need to know why. Not in a general sense. In a specific, client-call sense. If your spend jumped 40% last Tuesday, or if three campaigns went dark during a peak window, you need a sentence. Not a probability distribution. Not a model confidence score.
If an AI moved your bids last Tuesday, explaining that to a client is genuinely hard. The model saw a pattern, made a call, and the logic lives inside a weight matrix you have no access to. That is not a knock on the model. It may have been exactly right. But the explanation does not transfer. "The algorithm decided" is not a strategy your client signed off on, and it is not something you can point to in a reporting doc.
If a performance rule fired because ACoS hit 38% on three campaigns, that is a sentence. You wrote it. You can say it. "We have a rule that alerts us when ACoS crosses our target threshold. It fired on these three campaigns yesterday. Here is what we looked at, and here is what we decided to do next." That is an accountable workflow. That is a client relationship.
Rules do not drift
AI models optimize on historical patterns, and the account you have in September is not the account you had in March. Seasonality shifts. Creative rotates. A new competitor enters a category. The model that learned from Q1 data will keep making Q1 decisions until you retrain it or the vendor does, often without any visible signal that the optimization has gone stale.
Rules do not have that problem. A dayparting schedule you built in September runs the same way in December unless you change it. A budget rule that holds spend flat on Sundays holds spend flat on Sundays in week one and week forty. The logic is not adapting to signals you did not approve. You know exactly what will happen, because you said exactly what should happen.
That stability matters more than it sounds. Accounts managed by agencies are not experiments. They are ongoing commitments to a client's brand and budget, renewed every quarter or every year. Predictable behavior is a feature, not a consolation prize for a product that could not build a model.
Agencies have a different standard
When you are managing Amazon Ads across multiple clients, the risk profile changes entirely. You are not optimizing your own account on your own risk tolerance. You are making decisions in someone else's account, with someone else's money, against targets you agreed to hit in a proposal.
Rules let you draw bright lines. "We never pause campaigns during the 7pm to 9pm window for this brand." "We do not touch budget on days when a new product is launching." "Spend alerts go out the morning after, not in real time, so we can review before acting." Those are rules. You can write them down. You can put them in a kickoff deck. You can show a new team member exactly how the account behaves and why.
An algorithm does not know that Parkway Home has a Monday standup where the client reviews weekend spend, and you need that data clean before the call. It does not know that Harbor Kitchen's brand standards prohibit pausing their flagship campaign under any circumstances during a live promotion. Those constraints live in your head, in your agency playbook, in your client agreements. Rules are how you get them into the product.
What this is not
This is not an argument against AI. Off Hours uses AI to build rules. When you describe what you want in plain language, the AI builder translates that into rule logic. You review the output, adjust anything that looks off, and save it when you are satisfied. The AI is doing the writing. You are doing the deciding.
Northlane Goods uses natural-language input to draft their dayparting rules, then reviews the logic before activating anything. The rules run on their timeline, not the model's. That is the distinction we care about: AI as a writing assistant that speeds up your work, not as a decision maker that replaces your judgment.
The honest tradeoff
Rules require thinking upfront. You have to know your account well enough to write useful ones. What hours convert for you? What ACoS benchmarks should trigger a review? What does a healthy spend day look like versus a drift day? Those questions have answers, but you have to find them first.
The black-box tools promise to handle that discovery for you. Feed them enough data and they will figure it out. For some accounts and some operators, that works. We are not here to dismiss it.
But we think the accounts where you have done the thinking are the ones that compound over time. The team that knows their targets well enough to encode them as rules is the team that catches the anomaly before the client does, that onboards a new account faster because the logic is documented, and that walks into a QBR with a clear story about why the account is where it is. The thinking you did upfront pays out across every week the rules run.
Rules are how you build an account you understand. An account you can explain, hand off, and stand behind. That is the bet we made, and it is the product we built.
Off Hours is Amazon Ads automation you direct. Write the rules once, in plain language, and Off Hours runs them on every account, every day. Start a free 14-day trial.