AI Governance Requires Systematic Data and Human Oversight
Laerte Finotelli explores how balancing algorithmic suggestions with human context helps prevent biased decision-making.
Automating decisions with artificial intelligence agents requires robust governance and audit layers. In a discussion on Marketing - The Podcast, product coordinator Laerte Finotelli explains why technology should propose recommendations without replacing business judgment.
The rapid adoption of artificial intelligence agents across marketing and product workflows has brought data governance challenges to the forefront. In a conversation with Bruno Gimenes on Marketing - The Podcast, product coordinator Laerte Finotelli analyzed how automated decisions must balance systematic variable processing with human judgment rooted in business context.
In the video above, the full episode explores algorithmic governance and delves into product science, advertising metrics, and consumer attention. The dynamic between predictive modeling and decision-making becomes clearer when observing how algorithms correlate massive volumes of microdata to suggest actions.
The Logic of Microdata and Algorithmic Behavior
Discussing how technology processes information, Finotelli cites an analogy from Yuval Harari's book Nexus. While human reasoning often distrusts an excessive accumulation of isolated justifications for a single situation, machines operate precisely by cross-referencing dozens of simultaneous micro-signals to calculate a probability.
As an example of this dynamic, Finotelli points to surge pricing in ride-hailing apps. Parameters such as mobile battery levels, the likelihood of rain, exact geographic coordinates, or proximity to a concert venue are combined by the system to recommend a ride fare. The algorithm focuses on raw data correlation, unlike a person who would evaluate the situation through a different lens.
Audit Layers and Human Decision-Making
To prevent automated goals from creating distortions or running without transparency, Finotelli argues that technology should focus on generating evidence-based recommendations, leaving the final call to human judgment.
"I am strongly in favor of machines making suggestions. Then, we need personal governance steps to accept or reject that suggestion before making decisions," Finotelli explains, emphasizing the need to question the underlying correlations supporting every system recommendation.
On the other hand, automated processing offers the distinct advantage of mitigating purely subjective impressions and over-reliance on gut feeling. According to Finotelli, optimal governance operates across two complementary fronts: "it requires a systematic layer—evaluating how much stepping back as a human is beneficial to truly eliminate bias—paired with knowing where human intervention is needed to ground those data points in reality." This integration ensures that internal company policies, public regulations, and real-world awareness guide the ultimate outcome.