👋If you are a new reader, my name is Danar Mustafa. I write about product management focusing on AI, tech, business and agile management. You can visit my website here or visit my Linkedin here. I am based in Sweden and founder of AImognad.se – leading AI maturity Model Matrix. Get your free assessment here.

What Enterprise CIO’s think of Gen AI?

This a16z survey (June 10, 2025) captures insights from 100 CIOs across 15 industries—complemented by dozens of interviews—to map out how large enterprises are budgeting, procuring, and deploying generative AI at scale.

1. Budgets Exploding—From Experiment to Core Investment

Annual budget growth ~75%: CIOs describe the explosive spend with comments like, “what I spent in 2023 I now spend in a week”
GenAI graduating out of innovation budgets: Experimental funds once made up ~25% of AI spend; today that’s down to just 7%, with AI now financed by permanent IT and BU lines.
Why it matters: GenAI is no longer a pilot—it’s a central piece of business operations, warranting its own recurring budget allocation and long-term planning.

2. Multi‑Model Strategies Define Adoption

One model is no longer enough: A full 37% of CIOs now use five or more LLMs in production—up from 29% last year.
Model performance varies by use case: Anthropic for fluent writing, OpenAI for complex Q&A, Google for cost-efficient bulk inference—each model brings distinct strengths.
Smart orchestration wins: Enterprises are deploying model-agnostic “model gardens” that route tasks to the best-performing and most cost-effective model.

3. Closed-Source vs. Open-Source: A Delicate Balance

Open-source on the rise: While closed models still dominate overall, 46% of CIOs strongly prefer open-source, mainly for control and customization, not solely cost savings
Hybrid deployments are growing: Many enterprises self-host Llama, Mistral, and Meta models for privacy, while retaining closed models via cloud—blending flexibility with enterprise-grade support.

4. Procurement Matures: Enterprise Rigor Takes Over

GenAI decisions now mimic software procurement: CIOs conduct benchmark tests, security reviews, licensing negotiations, and align with SLAs—treating model providers like vendors such as SAP or Oracle.
Switching costs rising fast: As prompt structures and integrations proliferate, switching between providers grows both technically and economically painful.

5. Vendor Ecosystems Displace Custom Builds

Buy over build: More than 90% of enterprises are now deploying third-party AI apps (especially in customer service), instead of creating internal solutions from scratch.
CIO takeaway: Focus on verticalized or highly customized in-house apps. For generic use cases, mature third-party apps offer time-to-value advantages and built-in scale.

6. Talent & Infrastructure Remain High Priority

Models are 25% of the cost: The rest lies in implementation, fine-tuning, and Ops.
GenAI know-how is rare: Internal teams often lack integration and LLMops capability. Managed services from CSPs or boutique firms have become essential.

✅ Key Takeaways for Leaders & Founders

Insight	Implication
AI spend is strategic, not experimental	Target long‑term budgets, integrate deeply with core business ops
Multi-model flexibility matters	Orchestration between closed/open models can be a competitive edge
Build once, provide forever	Efficiency, control, vendor independence are key
Open-source momentum growing	Particularly for IP-sensitive and highly customized needs
Procurement resembles enterprise software	Strong sales, security, SLAs are table‑stakes
Resourcing is just as important as models	Support with services, tools, talent acquisition

✅ What CEOs & CIOs Should Do Today

Business Area	Action Items
Budgeting	Treat GenAI as foundational—build 50–100% YOY growth into budgets; integrate into IT and BU of P&L.
Model Strategy	Invest in multi-model orchestration; continuously benchmark performance, cost, and data governance.
Source Mix	Develop hybrid deployments: open-source self-hosting for privacy; closed-source via CSPs for reliability.
Procurement	Institute full due diligence: benchmarks, security, integration, legal/risk processes, and exit strategies.
Build vs Buy	Let third-party apps handle standard use cases; reserve internal development for bespoke or IP‑rich needs.
Talent & Ops	Budget for LLMops and integration costs; build internal talent via training or lean on external partners.