The uncomfortable conversation about AI and agencies got real when Tomer Tagrin posted publicly about sunsetting parts of the old model. He is right that AI is coming for a lot of agency work. He is also right that it will not take all of it.
What AI genuinely replaces
The commodity execution is going first: boilerplate code, first-draft copy, routine QA, the kind of task with a clear input and a predictable output. If your agency’s whole value is producing those deliverables faster than the client could, that value is eroding fast. Pretending otherwise is not a strategy.
I would rather name it than hide from it. An agency built purely on execution is exposed, the same way an agency that tries to be only a dev shop was already exposed before AI showed up.
What AI does not replace
Judgment, trust, and accountability do not automate. A client does not want ten AI-generated options, they want someone to tell them which one is right and to own the outcome. That is a relationship, not a deliverable. It is why human connection still beats AI marketing and why founder stories beat AI content. The parts buyers actually trust are the parts a model cannot fake.
Dedication is in that category too. The five clients who told me why they stay did not point at deliverables. They pointed at being cared for. No model does that.
What survives, in one line
The agencies that make it will use AI to do the commodity work cheaply and spend the freed-up time on the things only humans can do: strategy, honest advice, and standing behind the result. That is not a downgrade, it is a return to what a partner was always supposed to be. It is the same reason platforms change but the partner does not.
If you want a partner who uses AI as a tool and still answers for the outcome, book a call.