AI adoption isn't the same as deploying regular software or cloud apps. The most successful companies treat it as something completely new. OpenAI's report highlights seven key lessons from companies that are getting real results today.
Three Areas Where AI Delivers Results
Companies are seeing measurable improvements in:
- Helping people work better and faster
- Automating routine tasks
- Creating smarter, more responsive products
Lesson 1: Start with Testing (Evals)
Morgan Stanley began with rigorous testing of AI on specific tasks like translating documents and summarizing information. This careful approach built trust across the organization. Now 98% of their advisors use AI daily, document access jumped from 20% to 80%, and client follow-ups happen in hours instead of days.
Lesson 2: Embed AI in Your Products
Indeed uses AI to explain to job seekers why specific jobs match their skills. This personal touch boosted job applications by 20% and actual hiring success by 13%. With 20 million messages sent monthly, these small improvements create massive business impact.
Lesson 3: Start Now and Invest Early
Klarna built an AI assistant that now handles two-thirds of all customer service chats, cutting resolution time from 11 minutes to just 2. The project is expected to save $40 million. This didn't happen overnight - they improved through constant testing and refinement. Now 90% of employees use AI daily.
Lesson 4: Customize Your AI Models
Lowe's fine-tuned AI models for their specific product data. This improved product tagging accuracy by 20% and cut errors by 60%. Rather than using AI "out of the box," they tailored it to understand their unique product categories and customer search patterns.
Lesson 5: Get AI to Your Experts
BBVA bank put AI tools directly in the hands of employees who know their workflows best. In just five months, staff created over 2,900 custom AI tools. Their credit, legal, and customer service teams built solutions that turned week-long projects into hours-long tasks.
Lesson 6: Unblock Your Developers
Mercado Libre built an AI platform called Verdi that helps their 17,000 developers build better applications faster. This improved everything from inventory management to fraud detection (now 99% accurate) and localizing product descriptions for regional language differences.
Lesson 7: Set Bold Automation Goals
Even OpenAI automates their own work. They built a system to help support teams craft responses and take actions for customers. This handles hundreds of thousands of tasks monthly, freeing people to focus on more important work.
The Bottom Line
The companies seeing the biggest wins aren't racing to put AI everywhere. They focus on high-value, easy-win use cases first, learn what works, then expand. The results speak for themselves: faster proce