Despite the ongoing AI boom, enterprise-scale AI adoption remains sluggish.
The fifth annual UBS corporate AI survey led by Karl Keirstead reveals only 17% of respondent companies have achieved large-scale AI deployment, a marginal increase from 14% in March 2025.
Notably, 60% of enterprises prefer building AI solutions in-house rather than purchasing third-party products, while just 5% have successfully scaled "AI agent" implementations.
Market leaders Microsoft, OpenAI and Nvidia continue dominating enterprise AI. Microsoft Azure leads in cloud infrastructure, while OpenAI's GPT models claim three of the top five spots in large language models despite growing competition from Google Gemini and Anthropic Claude. Microsoft's M365 Copilot remains the preferred enterprise AI tool, though OpenAI's ChatGPT Business edition is rapidly gaining ground.
Conducted in October 2025 with 130 IT executives (average company size: 8,200 employees, $800M IT budget), the survey identifies key adoption barriers: unclear ROI (59%, up from 50% in March), compliance concerns (45%), and talent shortages (43%).
Contrary to automation fears, 40% of respondents expect AI to increase headcount versus 31% anticipating reductions, with only 1% predicting significant layoffs - positive news for seat-based SaaS providers.
The DIY AI Revolution Only 34% of enterprises rely solely on third-party AI software, while 60% pursue fully in-house or hybrid development approaches. This challenges traditional SaaS vendors' assumption that complexity would drive buyers toward off-the-shelf solutions.
UBS notes this trend creates opportunities for model providers like OpenAI and Anthropic to sell "model+tool" platforms to enterprises preferring custom builds.
In application scenarios, internal IT helpdesk AI adoption (75%) outpaces external customer support (52%). ServiceNow leads in IT workflow automation, surprisingly outperforming Salesforce in CRM AI rankings.
AI Agents: Hype vs Reality Despite being touted as the next frontier, only 5% of enterprises have deployed AI agents at scale, with 71% in pilot stages and 22% yet to begin testing. This slow adoption tempers expectations about AI agents displacing human labor and suggests vendor revenue projections may be overly optimistic, with meaningful adoption likely post-2027.
The Microsoft-OpenAI Stronghold OpenAI maintains enterprise LLM leadership with GPT 5.0, 4.0 and 3.5 occupying three top-five positions. Despite claims of LLM commoditization and Google's technical advances, OpenAI's market position remains strong - though competition intensifies as Google Gemini adoption surged from 19% to 46% year-over-year, with Anthropic Claude rising to third place.
In productivity tools, M365 Copilot leads but ChatGPT Business is closing the gap. Enterprises now average 2,050 Copilot seats (up 67% YoY) versus ~995 ChatGPT seats.
Data Infrastructure Winners AI drives disproportionate growth in data infrastructure spending, with 52% of respondents planning budget increases versus only 10% expecting cuts.
Cloud data warehouses show strongest momentum (69% planning increases, 25% significant boosts), benefiting Snowflake, AWS Redshift and Google BigQuery in a tightening race. Data lakes (56% growth) and ML/AIOps platforms (60%) also outperform operational databases (only 10% planning major increases), reflecting current AI focus on analytics over transactional applications.
