State of AI in Business 2025 — The GenAI Divide

Executive summary

Despite $30–40B in enterprise investment into GenAI, the report finds that 95% of organizations are getting zero return. We identify a pronounced GenAI Divide: while exploration and pilots are pervasive, only a small fraction of initiatives—about 5%—drive measurable P&L impact. The divide is characterized not by model quality or regulation, but by approach: winners focus on learning-capable, workflow-integrated systems; losers deploy static tools that fail to adapt.

1. Adoption vs. Transformation (Key Figures)

2. The GenAI Divide: causes and consequences

The primary barrier to scaling is the learning gap. Most deployed GenAI systems do not retain feedback, adapt to context, or improve over time. Additional factors include poor workflow integration, lack of trust among users, and investment bias favoring visible front-office use cases (sales & marketing) over often higher-ROI back-office automation.

A key behavioral pattern is the shadow AI economy: employees independently use consumer LLMs (ChatGPT, Claude, Copilot, etc.) to automate tasks—often achieving better ROI than sanctioned pilots. Roughly 90% of workers report personal AI use, while only ~40% of companies have purchased enterprise LLM subscriptions.

3. Where ROI actually appears

High-value ROI was most commonly observed in back-office functions (BPO elimination, document processing, finance & procurement) and in narrow workflow automations such as:

• Call summarization and routing
• Document automation for contracts and forms
• Code generation for repetitive engineering tasks

Documented impact examples: BPO elimination yielding $2–10M annually; 30% reduction in agency spend; 40% faster lead qualification in front-office scenarios.

4. Best practices to cross the divide

Organizations that succeed tend to:

• Prefer strategic partnerships (buy & co-develop) — external partnerships reached deployment ~67% of the time versus ~33% for internal builds in the interview sample.
• Empower line managers and prosumer power users to lead adoption instead of relying solely on central labs.
• Benchmark vendors on operational outcomes and hold them to business KPIs rather than model-only metrics.
• Focus on narrow, high-value workflows and continuous learning rather than broad, brittle solutions.

5. The road ahead — agentic systems & memory

The next wave of meaningful adoption depends on systems that maintain persistent memory, learn from feedback, and autonomously orchestrate complex workflows (termed agentic AI). Protocols and frameworks such as MCP, A2A, and NANDA are already shaping an "Agentic Web" where interoperable agents coordinate across systems. The report notes a narrow 18-month window in many verticals for organizations to lock in learning-capable vendor relationships.

Full report text (selected excerpts)

Executive Summary
Despite $30–40 billion in enterprise investment into GenAI, this report uncovers a surprising result in that 95% of organizations are getting zero return. The outcomes are so starkly divided across both buyers and builders that we call it the GenAI Divide. Just 5% of integrated AI pilots are extracting millions in value, while the vast majority remain stuck with no measurable P&L impact.

The core barrier to scaling is not infrastructure, regulation, or talent. It is learning. Most GenAI systems do not retain feedback, adapt to context, or improve over time.

(Report continues — full technical appendices, methodology, and interview protocol available on request.)