For AI Adopters, Success and Failure Look the Same Right Now
Exponential View (Azeem Azhar) · rss · 2026-07-30
Despite massive corporate investments in AI, broad productivity gains remain elusive. BCG surveys indicate half of CEOs globally feel their jobs depend on getting their AI strategy right, yet public disclosures of net AI returns remain rare.
Azeem Azhar argues that in the rollout of general-purpose technologies, the early signals of success and failure often look identical. Companies must navigate an investment "J-curve":
- High Learning Costs: The bills for trial-and-error, skill training, and organizational change arrive well before any returns.
- Adopter Archetypes: The essay uses historical case studies to illustrate two types of technology adopters:
- The Bounded Adopter: Finds something that works and stops experimenting. For example, Borders outsourced its e-commerce to Amazon, which worked short-term but destroyed its long-term capabilities, leading to bankruptcy.
- The Project Accumulator: Keeps launching new projects without learning from them. For instance, GM in the 1980s bet on multiple automation projects simultaneously; despite the success of its NUMMI joint venture with Toyota, the lessons failed to scale across the company.
The piece emphasizes that evaluating AI transformation requires looking beyond short-term ROI and focusing on whether a company is building mechanisms for continuous learning and knowledge accumulation.
More from AGI Musings
- LeCun: Industry AI Research Still Heavily Relies on the Academic Ecosystem — ylecun · 2026-07-30
- AI Companies Pivoting to Biomedical Research: Bullish for Humanity, Bearish for Business? — iskander · 2026-07-30
- LLMs Disrupt YouTube Tutorial Ecosystem: Indie Creators Pivot, Brands See Opportunity — misraturp · 2026-07-30
- 12 Terminal Pain Points in the Current AI Paradigm: From Cost to Amnesia — JimR_Ai_Research · 2026-07-30
- How Do Developers Distribute and Monetize Finished AI Agents? — FeeNational635 · 2026-07-30
- AI Safety Researchers Debate Slowdown Need and HuggingFace Incident Transparency — davidmanheim · 2026-07-30