The AI Adoption Gap Is Organizational, Not Technical

Mark Pincus sees AI as a waiting game: consumer apps are too expensive to scale now, but costs will drop. The real signal is that 90% of enterprises see no AI benefit—because they haven't changed processes or output expectations—making the bottleneck organizational, not technical.

His ‘proven, better, new’ product framework offers a way out. Copy what works legally (‘proven’), improve it by an obvious margin everyone would accept (‘better’), and test novel ideas fully expecting them to fail (‘new’). This separates disciplined execution from wishful innovation and forces teams to isolate what they're actually testing.

For engineering leaders, the implication is clear: token spend without workflow redesign is waste. Budgeting for AI should pair compute dollars with explicit changes in how teams ship and measure. The hype curve of ‘always-on AI assistants’ is seductive but currently unaffordable below enterprise price points.

The contrarian take: consumer's current ‘uninvestable’ status is exactly why forward-looking teams should experiment now, using expensive inference as a time machine for when compute becomes free. The risk is burning cash with no immediate return; the tradeoff is getting a head start on services that will feel inevitable later.