We stopped selling small businesses a self serve AI ads tool because marketing teams did not have time to run it themselves, and running it for them instead cut one account’s cost per enquiry from about €47 to €11.17 in eight weeks.
We built a piece of software that reads Google and Meta Ads accounts every day and finds what is wrong with them. Wasted spend, missed searches, keyword issues, and it suggests campaigns, makes optimisations, and builds out strategy off the back of real live data. It still works. Every account we run today is read by the same engine.
But it does require you to give it feedback, and correct its inevitable mistakes. The more you do this, the better it becomes, and the less you have to do it. But you still have to do it.
For a while we sold access to the SaaS product directly. A business could connect their account and get all the features.
But marketing teams at SMEs do not have the time to train an AI. Handing them a longer, more accurate list of problems is not the same as solving the problem. It can make things worse, because now the gap between what they know is wrong and what they have time to fix is visible, in writing, every week.
What they actually want is simpler than a dashboard. They want leads, at a cost that makes sense, without becoming an expert in Ads to get them.
Why AI still needs someone catching its mistakes
Good decisions from data still need a person who understands the business behind the numbers, and who catches it when it gets something wrong. Software can tell you a number moved. It cannot tell you a quiet fortnight was a planned closure, that a client needs bad news delivered gently this month, or that a competitor’s price change means your bid strategy is now wrong for a different reason than it was last week.
Every time it gets something wrong, someone has to correct it, and every correction is part of what makes it sharper next time. That feedback loop never fully closes. It gets lighter the longer you run it. It does not go away.
In a self serve model, that ongoing correction work fell to the business owner, on top of the job they already had. Most reasonably decided that training a piece of software was not it. They were happy to pay for the agency.
So in July, after months of testing the alternative properly, we stopped selling the software as something you trained yourself. Our own specialists took that job over, for every account we run. Same engine, reading every account across Google and Meta every day, suggesting campaigns and optimisations off live data. The difference is who corrects it when it is wrong and who feeds it what it cannot see on its own. Instead of a business owner or in house marketing team training a tool around everything else on their plate, a dedicated experienced Glitch strategist who already knows the account does that correcting daily, as part of running it, not as an extra task on top.
What changed on the accounts we run
The clearest way to show what changed is one account we have run this way since the summer.
In the eight weeks after the move, cost per enquiry on it fell from about €47 to €11.17, a drop of roughly three quarters. Return on ad spend went from 1.70x to 5.44x over the same period, while monthly spend on the account grew rather than shrank.
None of the individual fixes behind that were complicated. What changed was that someone was correcting the account’s course every day instead of once a month, because the software never stopped surfacing what needed a decision, and a specialist was already there to make it.
We are not arguing that the software matters less now. It matters more, because it is being fed real context and corrected every day rather than occasionally. What changed is who does that work. It was never really a job for the business owner to fit in around everything else. It was always going to need someone dedicated to it, and the fastest way to get a good result is to put that person between the software and the outcome, not hand over a tool and hope the correcting happens somehow.
That is the model now, the one we mean when we say AI native. The software suggests, optimises, and builds strategy off live data every day, doing the volume work no person could do daily at a price a small business would pay. A dedicated strategist corrects it, feeds it what it is missing, and answers for the result. Bringing the human context into play. Which is the game changer.
You pay one flat fee for both, not a percentage of your ad spend.
If you want to see what the same read would find in your own account, the audit is free, and it comes with a 90 day strategy in writing before you pay us anything.
A few questions worth answering directly
Does this mean the AI makes the decisions? No. It suggests and optimises from live data, but it makes mistakes, and someone has to catch and correct them. A strategist does that daily, because correcting it needs context about your business the software does not have.
Do I have to train the AI myself? Not anymore. You did in the self serve version, which is exactly what most small business marketing teams did not have time for. Now our specialists do that daily, on every account we run.
What happened to the self serve version? We stopped selling it in July, after testing this model properly. The engine behind it is the same one we use for every account we run today. What changed is who does the ongoing work of feeding it context and correcting its mistakes.
Does this cost more than a typical agency? No, the opposite in most cases. €499 a month, flat, covers up to €10,000 in monthly ad spend, with no percentage taken from what you spend.
Glitch Ads runs Google and Meta advertising for small and growing businesses across Ireland, the UK and the US.
If you’re an SME or a consultant who needs an experienced AdOps team to run your Ad campaigns, get in touch. Glitch is now officially an AI native agency, helping businesses with their websites, fixing conversion tracking, building out strategy, and launching campaigns that are set up for success from day one.



