Where companies are deploying AI

Most companies have adopted AI somewhere. Far fewer have changed how they work. Here's what's different between the two.

There's a version of "we use AI" that means someone on the team has a ChatGPT/Claude/Grok/Gemini subscription. And there's a version that means specific workflows have been rebuilt around what AI can actually do reliably. The gap between those two is where most of the interesting and difficult work is happening right now.

McKinsey's 2025 State of AI survey puts 88% of organizations in the first camp: using AI in at least one function. Only 7% have scaled it across the business. That’s a failure to move from tool adoption to workflow redesign, which turns out to be a much harder problem.

"64% of organizations are using AI agents to automate repetitive internal workflows: follow-ups, status updates, internal reports. Tasks that previously existed in the gap between 'someone's job' and 'nobody's priority.'"

-- Index.dev, "Enterprise AI Agent Research", 2025.

The functions where companies are seeing measurable returns, not just productivity scores but actual cost and output numbers, have become identifiable. We mapped eight of them below, based on the latest enterprise research.

Engineering: where the ROI is clearest

Software development became the proof point for AI in operations because the outputs are measurable: pull requests, test coverage, deployment frequency. 90% of software professionals now use AI tools regularly. Developers report 40-55% faster output in controlled studies, and the real-world numbers are following.

The market for AI coding agents (which can read entire codebases and execute multi-step tasks) went from near zero to $4 billion in enterprise spend in 2025.

The underrated: internal communication

Meeting notes and email don't sound strategic. But they eat an enormous share of knowledge worker time, and AI is making a measurable dent. 82% of professionals now use AI tools in their inbox daily. Law firms cut meeting-minutes production from two days to two hours. Utility companies reduced the creation of 1,000 standard operating procedures from one hour each to ten minutes.

These are cases that show positive ROI within 90 days, require no custom model training, and get adopted by teams voluntarily rather than by request.

• 40% average productivity boost reported by employees using AI regularly
• 27 hours saved per week by the heaviest AI users inside enterprise teams
• 3x faster revenue growth per employee in sectors with high AI exposure
• 7% of companies have actually scaled AI across the entire organization

The real barrier is redesigning workflows.

In 2024, of all the companies that evaluated enterprise-grade AI systems (not generic chatbots, but tools built to change how a specific process runs), 60% made it to evaluation, 20% reached pilot stage, and only 5% reached full production.

The reason most stopped isn't what we'd expect. It wasn't the cost, the model quality, or the lack of buy-in from management. It was simpler: they tried to plug AI into how they already worked, and got underwhelming results. AI added on top of a broken or outdated workflow just automates the inefficiency.

The companies that made it to production did something different. They looked at the workflow first, asked what actually needed to change, and then built AI into the new version of it. McKinsey's analysis of 200+ AI transformations confirms this: workflow redesign is the single strongest predictor of real business impact.

The tool is the same. The difference is whether it's embedded in how the work gets done, or just sitting next to it.

Our approach

We've done this internally. Over the past year, we've mapped and rebuilt workflows across departments at AccVentures: from how we run research and track portfolio signals, to how we handle content production, internal reporting, and cross-team communication. The productivity gains are real, but more importantly, it's changed what our team can take on without adding headcount.

We're now working closely with our portfolio companies to implement the same approach across their operations: identifying the workflows worth redesigning first, building the right integrations, and making AI part of how the team works.

Sources

1. McKinsey. The State of AI 2025: Agents, Innovation, and Transformation. mckinsey.com
2. McKinsey. Superagency in the Workplace: Empowering People to Unlock AI's Full Potential at Work. mckinsey.com
3. Menlo Ventures. 2025: The State of Generative AI in the Enterprise. menlovc.com
4. MIT Sloan / NANDA. The GenAI Divide: State of AI in Business 2025. mlq.ai
5. BCG. AI Adoption in 2024: 74% of Companies Struggle to Achieve and Scale Value. bcg.com
6. Microsoft. AI-Powered Success: 1,000+ Stories of Customer Transformation. microsoft.com
7. Superhuman. State of Productivity AI 2025. blog.superhuman.com
8. Index.dev. 50+ Key AI Agent Statistics and Adoption Trends in 2025. index.dev
9. Harvard Business School. Study on AI-Assisted Task Completion and Quality. hbs.edu
10. IBM Research. AI Automation and IT Productivity Survey, 2025. ibm.com
11. US Department of the Treasury. Annual Report to Congress on the Use of AI, FY2024. home.treasury.gov