UiPath's CMO Replaced Outside Contractors With an AI Agent for Advertising

UiPath's marketing team no longer pays outside agencies to collect advertising data. An AI agent now does that work. The company's head of paid media built it on UiPath's own platform. Chief marketing officer Michael Atalla described the setup to TNW in a podcast. According to him, the agent collects data, analyzes it, and makes low-cost changes to how ads run.

What was actually automated

This is the most tedious part of performance marketing. Someone has to export data from ad accounts, merge it into a single report, spot underperforming campaigns, and adjust the settings. For years, contractors handled these tasks: it was cheaper to keep an outside team than to pull in-house specialists away from other work. Now an agent covers the whole chain.

The phrase "low-cost changes" matters. It suggests the agent does not control large budgets and does not make strategic decisions. It works in an area where mistakes are cheap and changes can be rolled back. That is how the first successful business deployments of agents usually look: a narrow process, clear metrics, limited permissions.

There is one more detail: the agent was built not by an engineer but by the head of media buying. For the agentic automation market, that is a strong signal, because the main barrier there has long been not the model but who builds the workflow, and how.

Why the case is interesting and where it is debatable

UiPath grew out of Bucharest (the company was founded in 2005) and started with RPA, meaning software robots for routine operations. Today it sells agentic automation. So the "we use our own product" story partly works as a showcase. Skepticism is warranted here: the available excerpt gives no figure for the agency budgets saved, no data on how campaign performance changed, and no information on the agent's error rate.

Judging by its title, the interview goes beyond marketing. It covers AI "slop", meaning the flood of low-quality generated content, the fate of junior positions, and advice from a Shaolin monk. The available excerpt does not show what exactly Atalla said on these topics, and we won't speculate. But the connection is obvious: if an agent takes over routine work that used to go to outside contractors and early-career specialists, the question of how people enter the profession becomes a practical one.

What is known about the case

• Customer: UiPath's internal marketing team.
• Executor: an AI agent on the UiPath platform.
• Author: the head of paid media, not an outside vendor.
• Functions: data collection, analysis, low-cost adjustments to ad delivery.
• What was removed from the budget: payments to outside firms for preparing advertising data.
• Not disclosed: the size of the savings, the depth of the agent's access, the oversight mechanism.

What to take away

If your team has a contractor who compiles reports on a schedule, start there. Describe the process step by step and mark where a human makes the decision and where a rule is enough.

Give the agent permissions only for reversible actions: limits on changes to bids and budgets, a log of every edit, and manual approval for anything that crosses a threshold. Count not only the agency's cost but also the process owner's time spent maintaining the agent, or the savings will turn out to be illusory.

Think separately about people. If this routine work was a training ground for newcomers, work out in advance where they will build experience instead. Otherwise, in a couple of years, there won't be enough people who can check the agent's work.