AI in the Enterprise 2025: Five Trends Redefining How Businesses Invest, Build, and Buy

16 Changes to AI in the Enterprise: 2025 Edition

A detailed report from VC firm Andreessen Horowitz offers an in-depth look at how generative AI is being adopted across enterprises. It covers investment trends, model usage, vendor preferences, and shifts in acquisition strategies. Here are five key insights that stand out for product development professionals:

• Enterprise investment levels in AI
• Enterprise usage of AI models
• Evolution in AI technology acquisition
• Why buying AI agents is becoming more common than building them
• Reasons enterprises prefer AI-native startups

The findings are based on research involving 100 CIOs from 15 industries, providing a solid foundation to understand current enterprise AI practices.

AI Spending on the Rise

Spending on Large Language Models (LLMs) is accelerating. Enterprise leaders expect AI budgets to grow by an average of 75.7% from $7 million in May 2025 to $12.3 million in 2026. This follows a jump from $2.5 million in 2024. Growth is driven by broader use cases and increased employee adoption.

Notably, there's a shift toward applying AI in customer-facing scenarios, moving beyond internal efficiency gains. This signals that AI is becoming a strategic tool for external business value.

Another sign of maturity: 39% of firms now allocate GenAI spending under central IT budgets, up from 28% last year. Meanwhile, funding from innovation budgets dropped from 24% to just 7%, showing AI is moving from experimental to core operations.

Multiplicity of Models

Deploying multiple AI models is now a common practice. In 2025, 37% of enterprises reported using five or more models in production, up from 29% the previous year. This approach helps avoid vendor lock-in and allows selecting the best model for each use case.

This trend reflects growing confidence in evaluating and managing diverse AI tools to meet specific business needs.

AI Tech Acquisition Gains Enterprise Qualities

Enterprises are applying traditional software buying disciplines to AI solutions. Price controls and functionality evaluations have become more stringent. The importance of total cost of ownership has increased significantly compared to last year.

This shift shows that AI is no longer an experimental technology but one that requires careful procurement and governance.

Buying Not Building

Early in AI adoption, many enterprises built their own AI applications directly on models. The latest data reveals a clear shift toward purchasing third-party AI applications due to the growing ecosystem of AI apps.

Many companies find internally developed tools hard to maintain and insufficient for gaining a competitive edge. For example, over 90% of respondents testing AI for customer support prefer third-party apps. Software development teams also heavily rely on third-party AI tools for production and testing.

AI-Native Innovators

AI-native startups often outpace established vendors in product development speed and AI functionality. Enterprises cite innovation, native AI software (not retrofitted), and vendor flexibility as top reasons to choose AI-native firms.

An example is AI coding tools: Users of Cursor, a native AI code editor, show higher satisfaction compared to GitHub Copilot users. Cursor has a Net Promoter Score (NPS) of 67, while GitHub Copilot scores 50 among users who haven't tried Cursor, dropping to 24 among those who have used both.

AI Agent & Copilot Summit

The AI Agent & Copilot Summit is an event focused on AI applications like Microsoft Copilot and AI agents. Following its success in 2025, the 2026 summit will take place March 17-19 in San Diego. It’s a good opportunity for product teams to explore AI-driven tools and strategies.

For those interested in expanding AI skills relevant to product development, exploring comprehensive training resources can be valuable. Consider checking out Complete AI Training’s latest AI courses for practical, hands-on learning.