Claim
The primary reason companies fail to accelerate with AI is not a tool shortfall. Command-and-control structures that gate information and decisions by title create bottlenecks that AI tools cannot bypass.
Mechanism
In hierarchical organizations, context must travel up and down the reporting chain before action can be taken. An employee with access to excellent AI tools but no decision-making context cannot move faster. The bottleneck is structural: when authority and information live at the top, adding AI to the people at the bottom does not change what they are authorized to do or know. Verna distinguishes agents (deployed) from agency (granted). Pushing AI tooling into teams operating in bureaucratic cultures produces high adoption rates and flat output metrics, because the constraint was never the tool.
Conditions
Holds when: the organization has significant information hierarchies or approval layers. Most large and mid-sized companies fit this description. Symptoms include decisions that require manager sign-off, context that is siloed by team or title, and projects that move at the pace of the slowest stakeholder chain.
Fails when: the company already operates with flat structure and high information access for individual contributors. In those environments, better tooling does directly accelerate velocity.
Evidence
Verna writes from direct advisory experience with B2B SaaS companies:
"If you ungate access to information and encourage autonomy, suddenly... the context doesn't have to pass up-and-down the chain."
"Rapid innovation cannot happen in that kind of environment. Super high-impact individuals and high-velocity teams will shrivel up and die."
"The goal is not more agents. It's more agency."
She recommends testing the model with R&D or pilot teams before attempting org-wide transformation, noting that high-agency individuals may need to be recruited because organizational "muscle memory" prevents existing staff from embracing autonomy even when it is offered.
Signals
- High-performing individual contributors who joined for speed leave within 12 months citing slow decision-making.
- AI tool rollouts show high adoption rates but flat output metrics or unchanged cycle times.
- Projects requiring fast iteration get routed through the same approval chains that slow non-AI work.
Counter-evidence
The tooling-first camp argues that sufficiently agentic systems reduce the need for organizational authority. If agents make actions reversible and auditable, teams can move fast without requesting permission because the agent verifies its own work before committing. Verna does not address this mechanism.
Cross-references
- The bottleneck for AI adoption has moved from model capability to organizational capability (Dan Shipper on organizational capability as the AI adoption ceiling, June 2026)