Highlights of the "State of Enterprise AI Report from OpenAI"

Enterprise AI adoption is scaling fast and embedding into day-to-day work. Over the past year, weekly ChatGPT Enterprise messages grew about 8x, with average workers sending 30% more messages, and seats increased ~9x year-over-year. Custom GPTs and Projects have become core workflow tools: weekly users rose 19x year-to-date and about 20% of all Enterprise messages now flow through these tailored assistants and projects.

Developer/API adoption is also surging: more than 9,000 organizations have processed over 10 billion tokens, nearly 200 surpassed 1 trillion, and average reasoning token consumption per organization rose ~320x in 12 months - evidence that higher-intelligence models are being integrated into products and services. In the last six weeks, Codex usage shows deeper software development integration, with weekly active users up 2x and weekly messages up ~50%.

Workers report clear, measurable gains. Seventy-five percent say AI improves speed or quality; typical users save 40–60 minutes per active day, and data science, engineering, and communications save 60–80 minutes. Function-specific improvements are broad: IT issue resolution is faster (87%), marketing/product campaigns execute faster (85%), HR engagement improves (75%), and engineers deliver code faster (73%). AI is also expanding who does technical work: 75% of workers say they can now perform new tasks (coding, spreadsheet automation, tool troubleshooting, agent/GPT design), and coding-related messages outside engineering/IT/research rose 36% over six months. Impact scales with intensity of use—workers who save >10 hours/week consume far more “intelligence” (credits) and use more models/tools across more task types.

Adoption is broad-based across industries and geographies, with uneven depth. The median sector grew >6x year-over-year; technology led with ~11x growth, while healthcare and manufacturing are among the fastest risers from a smaller base. Professional services, finance, and technology currently have the largest absolute ChatGPT Enterprise footprints. API usage is diversifying beyond tech: customer service and content generation now make up ~20% of activity, and non-tech firm API use grew 5x. International growth is accelerating: Australia (+187%), Brazil (+161%), the Netherlands (+153%), and France (+146%) grew faster than the global average; the U.S., Germany, and Japan remain among the most active by message volume, and Japan is the largest corporate API market outside the U.S.

A widening performance gap is emerging. Frontier workers (95th percentile) send 6x more messages than median workers and use advanced tools much more intensively (e.g., 16x in data analysis among analytics workers; coding sees a 17x message gap). Using AI across more distinct task types correlates with dramatically higher time savings. At the firm level, frontier enterprises generate ~2x more messages per seat and ~7x more messages to GPTs than the median, indicating deeper workflow standardization and organizational integration. Even among monthly active ChatGPT Enterprise users, a meaningful share has never touched some of the most capable tools (19% haven’t used data analysis; 14% haven’t used reasoning; 12% haven’t used search), underscoring headroom.

Case evidence points to business outcomes beyond productivity. Intercom used OpenAI’s Realtime API to cut latency by 48% for voice agents and now resolves about 53% of calls end-to-end; escalated calls complete 40% faster, translating to significant support cost avoidance. Lowe’s Mylow and Mylow Companion now answer nearly 1 million questions/month, more than doubling online conversion rates and lifting in-aisle customer satisfaction by 200 basis points. Indeed’s AI matching plus GPT explanations increased started applications by 20% and downstream interviews/hiring by 13%; Career Scout helped job seekers apply 7x faster and be 38% more likely to be hired (84% rating it valuable). BBVA’s legal chatbot automates ~9,000 queries annually, redeploys ~3 FTEs, and delivers 26% of its Legal Services savings KPI. Oscar Health’s integrated member chatbots instantly answer 58% of benefits questions and handle 39% of benefits messages without human escalation. Moderna compressed a core analytical step in Target Product Profile drafting from weeks to hours, accelerating cross-functional alignment and decision quality.

OpenAI’s synthesis highlights patterns among leading firms: deep system integration (secure connectors into core tools), workflow standardization and reuse (Custom GPTs, API-powered assistants), executive sponsorship, data readiness and continuous evaluations, and deliberate change management via centralized governance plus embedded champions. The report’s conclusion is clear: depth of usage-especially of advanced capabilities like reasoning, data analysis, Custom GPTs/Projects, and APIs—drives larger productivity gains, broader task coverage, and increasing revenue/customer experience impacts as AI shifts from ad-hoc assistance to core infrastructure.

Strengths

• Scale and diversity: The analysis draws on aggregated usage from over 1 million business customers, plus a survey of ~9,000 workers across ~100 enterprises, offering a broad, real-world view of adoption patterns and impact.
• Mixed evidence types: Combining telemetry (messages, tokens, tool usage) with worker-reported outcomes and concrete case studies yields a rounded picture of how AI is embedded and where value shows up.
• Clear adoption dynamics: The report identifies the deepening of workflow integration (Custom GPTs/Projects), the rise of higher-intelligence usage (reasoning tokens up ~320x), and the uneven intensity between frontier and median users/firms—all actionable for operators.

Limitations

• Vendor-centric lens: Findings reflect OpenAI customers and products; cross-model or cross-vendor comparisons aren’t presented, which may overstate generalizability or omit alternative adoption routes.
• Self-reported productivity: Time-savings estimates come from surveys; while directionally useful, they may be subject to recall or optimism bias and don’t uniformly translate to realized business value without rigorous measurement.
• Early-stage causality: Many business outcomes are correlational and case-study based; broader, longitudinal controls would clarify causation and durability across contexts.
• Coverage gaps: High-level sector/geography figures mask intra-sector variance, regulatory constraints (e.g., healthcare/finance data governance), and organization-specific readiness factors that influence pace and realized ROI.

The report credibly shows enterprise AI crossing from experimentation to scaled, embedded workflows with measurable productivity and customer-impact gains, while revealing a significant maturity gap. The most substantial benefits accrue to organizations and workers that standardize workflows, integrate secure data access, and lean into advanced capabilities (reasoning, analysis, custom assistants, API). The primary constraint is shifting from model/tooling maturity to organizational readiness—systems, skills, governance, and change management.

Read the full report here

----

Newnex is a verified institutional VC network enabling direct syndication and co-investment with other VCs - no intermediaries and no SPVs.

Sign up free.www.newnex.io

#OpenAI #artificialintelligence #AI