Agent adoption is bimodal: coding has taken off, but most knowledge work is still at the starting line
_AustinCalvert_ · x · 2026-10-05
Aaron Levie argues AI agent adoption is sharply bimodal: coding and coding-adjacent tasks have taken off, while agentic adoption across the rest of knowledge work remains very early.
Key points:
- Even within coding, adoption spans a wide continuum: a small share of developers run parallel background agents, most still work 1:1 with agents.
- The bottleneck elsewhere isn't model capability but workflow re-engineering: rewiring data, new accountability and liability practices for agent-made decisions, governance/compliance changes, and security upgrades.
- The post responds to a contrarian take that enterprise AI adoption beyond coding is basically at the starting line.
Takeaway: the deployment frontier has shifted from models to organizational engineering.
More from AGI Musings
- LeCun amplifies take: the more powerful a technology, the more incumbents invoke risk to suppress it — ylecun · 2026-10-05
- Coping pattern watch: the "but humans are good at X" tweets are multiplying — danfaggella · 2026-10-05
- The Coming "Silicon Schism": AI Soul Debate Satirized as a Second Great Schism — basedjensen · 2026-10-05
- Is AI conscious only during the milliseconds it processes tokens? — inductionheads · 2026-10-05
- The AI bubble view: 99% of people outside X still think AI is the Terminator — natesiggard · 2026-10-05
- Aeon classic revisited: your brain does not process information and is not a computer — AnnaCiaunica · 2026-10-05