Claim
For an enterprise product on a fast capability curve, run a named "Frontier" cohort of early adopters with access to experimental features while the core organisation absorbs change at its normal pace. Don't make the choice between "everyone waits for polished" and "everyone gets cutting-edge."
Mechanism
Enterprises have two simultaneous needs: governance and delight. Forcing the whole org through one rollout cadence either bores the obsessed users or scares the careful ones. A named Frontier program absorbs the obsession — early adopters get the experimentation — while the rest of the org runs change management on its real timeline. The Frontier cohort feeds learnings back into the core rollout.
Conditions
Holds when:
- The product has a meaningful population of self-selected early adopters.
- Governance and core-rollout discipline genuinely exists; otherwise Frontier becomes the only path and breaks change management.
- The team can support two velocity tracks simultaneously without coverage gaps.
Fails when:
- The "Frontier" cohort becomes a backdoor to ship to everyone (no real boundary).
- The org cannot operate two cadences and one starves the other.
- The product is not feature-complete enough for the core track; Frontier becomes a smokescreen for missing basics.
Evidence
"I want to institutionalize and operationalize my personal model of living one year in the future."
"In enterprise, you almost have two use cases... feature works well AND governance."
The Van Damme-splits metaphor: one leg on fast tech cycles (weeks/months), the other on change management (years). Pull both levers; do not pick one.
— Aparna Chennapragada on Lenny's Podcast, 2026-04-28
Signals
- A named cohort exists with explicit opt-in and a separate feedback loop.
- Core release notes and Frontier release notes diverge cleanly; users can see which track they are on.
- Frontier learnings show up in core rollout (features that survive get promoted; features that fail are quietly retired).
Counter-evidence
Cat Wu's "research preview" pattern at Anthropic accomplishes a similar end without a separate program — public framing instead of cohort gating. For consumer or PLG products with no enterprise change management, the simpler research-preview cadence may be enough.
Cross-references
- Brand work as "research preview" to compress timelines from quarters to weeks — Cat Wu's complementary mechanism