AI Strategy Roadmap: The 30/60/90 Day Plan

Why 90 Days Is the Right Horizon for an AI Roadmap

In short

Ninety days is faster than annual planning, so it forces decisions, but long enough for real-world learning to surface. It compresses the planning-to-evidence loop to a single quarter, which is the cadence most executive teams already run on.

Annual AI strategy decks age badly. By month six, the model landscape has shifted, the use cases are stale, and the budget is half-spent.

A 90-day horizon forces a different posture. The team must pick one or two bets, scope them tightly, and produce evidence by the end of the quarter.

Ninety days is also long enough to be honest. A 30-day plan cannot include production deployment. A 12-month plan cannot avoid politics. Ninety days is the smallest unit of real learning.

BCG and MIT (2024) found that only about 26% of GenAI investments deliver the value expected. The gap is rarely technology — it is the gap between planning ambition and execution discipline. Ninety days closes that gap.

The 30/60/90 Structure: Three Milestones, Three Deliverables, Three Checkpoints

In short

Each 30-day block has exactly one milestone, one deliverable, and one stakeholder checkpoint. The constraint forces focus and gives leadership a predictable rhythm to govern the programme.

The 30/60/90 framework is not a list of activities. It is a cadence of three decisions.

Days 1-30: Readiness and scope. Milestone: pilot scope signed. Deliverable: one-page pilot brief. Checkpoint: executive sponsor go/no-go.

Days 31-60: Pilot in production. Milestone: working end-to-end flow on real data. Deliverable: validated pilot meeting KPIs. Checkpoint: governance committee review.

Days 61-90: Measurable ROI and scale. Milestone: production deployment plus ROI report. Deliverable: 12-month scaling plan. Checkpoint: board-level sponsor review.

The structure is deliberately spare. Add a fourth deliverable per phase and the roadmap collapses into a wishlist.

Source: Alice Labs 30/60/90 Roadmap framework

Common Pitfalls — and How to Avoid Them

In short

RAND Corporation's 2024 analysis of AI failure names three top causes: missing business owner, weak data foundation, and unclear metrics. All three are addressable in the first 30 days if the roadmap forces the right checkpoints.

RAND Corporation (RR-A2680-1, August 2024) analysed root causes of AI project failure. Three patterns dominate.

Missing business owner. When the project sponsor is IT or a centre of excellence, accountability for the business outcome is diffuse. The fix: in week 1, name a single business owner who controls a P&L line the pilot will move.

Weak data foundation. Teams discover in week 6 that the data they planned to use is incomplete, unlabelled, or behind a system boundary they cannot cross. The fix: in week 1, validate that the baseline data exists and is usable before committing scope.

Unclear metrics. The pilot ships, but no one agrees whether it worked. The fix: in week 3, write the metric and the target before code is written. Numbers, not adjectives.

A fourth pitfall is specific to 2026: skipping EU AI Act classification at the scoping stage. If the use case turns out to be high-risk under Annex III, governance work that should have started in week 4 surfaces in week 11 — and the launch slips.

Alice Labs Cases Mapped to the 30/60/90 Timeline

In short

Real engagements from Ljusgårda, public sector, and media show how the 30/60/90 cadence plays out across very different use cases — and why the Alice Labs Implementation Index 2026 reports a 14-week median.

The 30/60/90 cadence holds across industries. Three Alice Labs engagements illustrate the pattern.

Ljusgårda (consumer / cannabis). Days 1-30: readiness assessment identified content operations as the highest-ROI workflow. Days 31-60: AI-assisted content pipeline in production with measurable throughput gain. Days 61-90: ROI report and second pilot scoped for customer service.

Public sector engagement. Days 1-30: readiness assessment plus full EU AI Act Annex III classification (high-risk pathway). Days 31-60: pilot built with governance, logging, and FRIA running in parallel. Days 61-90: production launch with regulator-ready documentation.

Nordic media client. Days 1-30: shortlisted three candidate use cases, picked one with clear baseline cost. Days 31-60: pilot validated against editorial throughput metrics. Days 61-90: scaled to second newsroom and produced a 12-month plan covering five additional use cases.

Across 100+ Nordic engagements the Alice Labs Implementation Index 2026 reports a 96% production rate and a median 14 weeks from pilot start to production. The 30/60/90 roadmap is the framework that produces those numbers.

From 90-Day Pilot to 12-Month Roadmap

In short

The 30/60/90 framework is not a one-off — it is the repeating heartbeat of a maturing AI programme. After the first cycle, the next 30/60/90 launches inside the broader 12-month roadmap, with platform and governance investments compounding across pilots.

The first 30/60/90 produces evidence. The second produces momentum. By the fourth, the cadence is the operating system of the AI programme.

The 12-month roadmap is built from the 90-day learnings, not from a top-down vision deck. Each quarter, the leadership team decides which pilots scale, which new ones launch, and which platform investments unlock the next batch.

By month nine, two patterns appear. First, the cost per pilot drops as governance, platform, and data work compounds. Second, the team starts running pilots in parallel — one 30/60/90 cycle does not block the next.

By month twelve, the question shifts. It is no longer "can we do AI?" It becomes "which AI bets give the next quarter's return?" That is the moment the programme has matured.