You’ve probably encountered the headlines of companies struggling to see financial returns from their AI investments or executives cancelling AI rollouts due to disappointing ROI, and skepticism around AI’s value in the enterprise. Yet, paradoxically, if you’ve personally used AI tools, you’ve likely felt their value. Why is there this disconnect?
The Disconnect Explained
Imagine a hypothetical company with three employees:
- Alice (Product): Highly skilled in product management, but less so in engineering and marketing.
- Bob (Engineering): Expert engineer but weaker in product and marketing.
- Eve (Marketing): Expert in marketing, with minimal skills in product and engineering.
Collectively, the company’s competencies are strong across product, engineering, and marketing because each employee excels in their specialty. But what happens when we introduce AI?
Research shows that AI disproportionately boosts the productivity of non-experts, transforming them into “quasi-experts”. In our hypothetical company, that means Alice, Bob, and Eve each become significantly better in areas outside their core expertise. For example, Alice can now handle engineering tasks reasonably well, Bob becomes decent at marketing, and Eve can engage effectively with product tasks.
Individually, this represents massive gains in versatility but most organizations are structured by speciality. Alice’s newfound engineering skill or Bob’s marketing insight is lost if their roles and workflows remain rigidly segmented. The gains in individual productivity may not translate to organizational ROI without changes to the company’s structure.
Lessons from the Past
This scenario mirrors past historical technology transitions. For example, in transition from steam power to electricity, it took factories nearly 50 years to recognize the productivity gains. It took decades before factory owners realized they needed to reorganize completely, breaking away from centralized power sources of steam-power and redesign their workflows to fully leverage electricity’s potential.
Similarly, to capture the full value of AI, organizations must reconsider and possibly reinvent their structures and workflows.
The Frontier Firm, The Frontier Job
Microsoft’s 2025 Work Trend Index provides a glimpse into this future, introducing the concept of “Frontier Firms,” organizations where humans initially leverage AI assistants but eventually lead autonomous teams of AI agents.
The startup Oleve is an illustrative example. As discussed on the Latent Space podcast, this small company is generating outsized recurring revenue by empowering “product engineers” who blend multiple skillsets and enabling by AI agents to tackle tasks from product discovery to marketing.
AI has the potential to redefine your organization, your market and even the problem you’re solving for your customers.
For organizations seeking meaningful ROI from AI, the question isn’t merely “How can AI speed up our existing processes?” Instead, leaders should ask, “How must our processes evolve because of AI?”
Recommended Reading
Power to the people: How LLMs flip the script on technology diffusion
The many fallacies of ‘AI won’t take your job, but someone using AI will’