Executive: Run AI Pilots While You Clean Up Data

Businesses can speed up AI adoption by testing hypotheses at the same time as they prepare their data, according to Sergei Golitsyn, head of the IT holding T1.

Speaking to NEWS.ru, Golitsyn said pilot AI projects can be launched without extensive groundwork, since information inside the corporate perimeter is already cleaned and well structured. Data quality only has to be brought up to standard once AI solutions are scaled and models are trained, he explained.

Untreated data is good enough to test hypotheses without heavy spending, Golitsyn added. Investment in better data quality and labelling makes sense after results are confirmed.

He also noted that preparing data to train AI calls for different approaches and considerably larger investment, with foundational models especially dependent on data quality.