Artificial intelligence adoption is moving beyond pilot programs and entering mainstream business operations, according to recent research. While a significant majority of companies are deploying custom AI solutions (68%) and investing heavily in the technology (81% spending over $1 million annually), challenges related to data and infrastructure are hindering progress.
The study highlights the increasing maturity of AI implementations. Most companies now have dedicated AI leadership roles (86%), and investment is significant, with a quarter spending over $10 million per year. Popular applications include chatbots, virtual assistants, and increasingly, more technical applications like software development (54%) and predictive analytics (52%). Generative AI, powered by models like Google’s Gemini, OpenAI’s GPT-4, and emerging options such as DeepSeek and Llama, is also a key priority (57%).
However, the journey to successful AI deployment is proving more complex than initially anticipated. Over half of business leaders find model training more difficult than expected, and data quality remains a major impediment. Nearly 70% of organizations report AI project delays due to data-related problems.
Concerns around data security and sovereignty are driving a trend towards on-premises and hybrid AI infrastructure. Two-thirds of business leaders believe these non-cloud deployments offer better security and efficiency, and 67% plan to move AI training data to such environments. Data sovereignty is cited as a top priority by 83%.
While executives express confidence in their AI governance frameworks (around 90%), practical challenges in data labeling, model training, and validation persist. Talent shortages and integration complexities with legacy systems also contribute to project delays.
The shift to on-premises and hybrid solutions signifies a focus on control, security, and governance as AI adoption expands. For continued success, organizations must prioritize transparency, traceability, and trust in their AI initiatives.