---
id: ins_seasons-not-roadmaps
operator: Asha Sharma
operator_role: CVP, Microsoft AI Platform
source_url: https://www.lennysnewsletter.com/p/how-80000-companies-build-with-ai-asha-sharma
source_type: podcast
source_title: Asha Sharma — Product as organism, post-training, agentic society — Lenny's Podcast
source_date: 2026-04-28
captured_date: 2026-05-01
domain: [product, ai-native, leadership]
lifecycle: [strategy-bets, process-cadence]
maturity: frontier
artifact_class: framework
score: { originality: 4, specificity: 4, evidence: 3, transferability: 5, source: 5 }
tier: B
related: [ins_product-as-organism]
raw_ref: raw/podcasts/asha-sharma--product-as-organism--2026-04-28.md
---

# Plan in seasons keyed to secular changes, not 6-month roadmaps

## Claim
Six-month roadmaps die on contact with model launches. Replace them with seasons, periods defined by a set of secular changes (e.g., "season of agents," "season of reasoning models"). Seasons can last 3 to 12 months depending on capability waves. Within a season, set 4–6 week goals that ladder up; reserve slack for the slope.

## Mechanism
A traditional roadmap commits the team to a fixed plan for a fixed duration. AI capability shifts invalidate that plan mid-quarter. Seasons commit instead to a *thesis* about the current wave, what's changing, what customer problems matter now, what winning looks like. Tactics underneath are revisable; the thesis is the stable artifact. The team gets shared orientation without locked-in deliverables.

## Conditions
Holds when:
- Leadership is comfortable with looser commitment language at the top of the planning hierarchy.
- The team can articulate the season thesis clearly enough that 4-6 week goals roll up to it.

Fails when:
- The org is scaled and dependencies (legal, finance, sales) require firm dates. Seasons feel like cop-outs to those audiences.
- The thesis is vacuous ("season of growth"). Without a substantive secular-change identification, seasons devolve into theatre.

## Evidence
> "How does anyone plan a roadmap when GPT-5 just came out? We think about it as: what season are we in? Season one was prototyping. Then models and reasoning. Now agents. A season is denoted by a set of secular changes. Can last 6 months, 3 months, a year. We ground everyone on: what changes are we in? What customer problems? What's winning look like? That shared sense is north star."

· Asha Sharma on Lenny's Podcast, 2026-04-28

## Signals
- Quarterly OKRs are loose; 4–6 week goals are sharp.
- Annual planning produces a thesis paragraph, not a Gantt chart.
- The team can answer "what season are we in?" in one sentence.

## Counter-evidence
Brian Halligan and other large-public-company operators argue the Gantt chart is irreducible past a certain scale because the org cannot coordinate without firm dates. Seasons are a small/mid-cap pattern. Hybrid approach: seasons at the top, traditional roadmap underneath for deliverables that need them.

## Cross-references
- `ins_product-as-organism`, the underlying frame that requires this planning shift