BTG Pactual Technology Director Warns Against AI Projects Lacking Direction
November 2, 2026 · 2 min read
Gerson Júnior, the technology director at BTG Pactual, stated that speed alone is insufficient and direction is necessary for artificial intelligence projects. Speaking at the IT Forum Na Mata HPE, the executive argued that AI initiatives without clear direction often become costs rather than assets.
Referring to Gartner research, Gerson Júnior noted that half of generative AI projects are abandoned after the proof of concept stage. He attributed this failure to unusable data, unsustainable premises, and the automation of processes that should not exist. He emphasized that companies must identify the specific problem a project solves before implementation, warning that justifying token expenditures will be difficult if no actual problem is resolved.
Data quality was highlighted as a critical failure point. Gartner studies cited by the director indicate that 63 percent of companies are unsure if their data is ready for AI, and only 4 percent consider themselves prepared. In infrastructure and operations, 38 percent of failures are linked to unavailable or poor data. Gerson Júnior shared that at BTG Pactual, the rule is to disable any alert that does not generate an action within two days.
Comparing AI to racing telemetry, Gerson Júnior explained that while his Mercedes A45 monitors over 80 parameters ten times per second, the focus remains on a few key signals to increase speed. He illustrated the danger of bad data with an example of a broken sensor that reported zero degrees in 50 degree heat, causing the car to lose performance. He advised favoring two smaller projects with compatible data over one giant project that never finishes.
Regarding infrastructure, Gartner reports that only 28 percent of AI projects in infrastructure and operations achieve expected returns, despite 45 percent of AI investments being directed toward infrastructure. Gerson Júnior compared the infrastructure area to water utility services, where the system is only noticed when it fails.
On the human element, Gerson Júnior cited Microsoft research showing 80 percent of professionals lack the time or energy to produce more, while 82 percent of leaders plan to expand team capacity using digital agents. He cautioned against measuring AI benefits by the number of employees cut, noting a Gartner prediction that 30 percent of professionals replaced by AI will be rehired by 2029 because humans provide essential context and history.
Finally, he referenced an MIT study on learning. The study found that 89 percent of people who wrote texts without tools could remember and cite their writing, compared to only 17 percent of those who used AI. Gerson Júnior concluded that delegating tasks to AI impoverishes learning, whereas using it as a tutor can accelerate it.