Artificial Intelligence in Financial Services and Investment Activity: Evidence on Adoption, Performance, and Systemic Risk

Artificial intelligence (AI) has moved from isolated pilots to a core component of financial services and investment management. This paper synthesizes evidence published between October 2024 and November 2026 to answer three questions: how widely AI is adopted, what measurable outcomes it delivers in investment activity, and which systemic risks it creates. We apply a structured documentary analysis to regulator surveys, international-organization reports, industry surveys, and recent peer-reviewed and preprint studies. Results show that adoption is now majority-level (75% of surveyed UK firms in 2024 and 81% of surveyed firms globally in 2026), but advanced, institution-wide integration remains much lower, and supervisors lag industry. Large language model (LLM) based portfolio studies report promising yet regime-dependent performance: gains in stable markets often weaken in volatile ones, and hybrid LLM–quantitative designs are the most robust. Reported benefits rest largely on self-reported surveys, while risks concentrate in third-party dependence, model correlation, data privacy, and monitoring gaps. We conclude that AI is a productivity technology with uneven, not yet transformational, effects, and that investment use cases require human oversight, time-safe evaluation, and supervisory data collection.