Stop Writing 3-Year AI Strategies. Do This Instead.

I watched a Fortune 500 bank spend $1.2M on a three-year AI strategy document in 2023. By the time the board approved it, GPT-4 had made half the assumptions obsolete. The team spent the next four months rewriting it. Then Claude 2 dropped. More rewrites. That document never produced a single deployment. It produced meetings.

Why Traditional Strategy Timelines Collapse Under AI

Enterprise planning evolved around technologies that moved in 18-to-24-month cycles. ERP implementations. Cloud migrations. Core banking platform swaps. You could reasonably predict what the landscape looked like in year three because the tools barely changed between kickoff and go-live.

AI moves on a different clock. In 2024 alone, we saw GPT-4 Turbo, Claude 3, Gemini 1.5, open-weight models crossing the quality threshold for production use, and multimodal capabilities going from demo to deployment-ready. Any strategy written in January was partially wrong by March.

The average enterprise AI strategy document takes 4-6 months to produce. Stakeholder interviews. Vendor assessments. Capability mapping. Executive alignment sessions. By the time you get signatures, the technology assumptions underneath your recommendations have shifted. You are building on sand and calling it a foundation.

This does not mean you stop planning. It means you stop planning the wrong way.

The Rolling 6-Quarter Model

Here is what actually works in enterprise AI planning. I call it the Rolling 6-Quarter Model. You plan six quarters out, but with radically different levels of detail at each horizon.

Quarter 1 and 2: Execution detail. Specific projects, named owners, defined budgets, measurable outcomes. These are commitments. You know what models you are using, what data pipelines feed them, and what success looks like in production metrics.

Quarter 3 and 4: Directional bets. You know the problem space. You know the business case range. You have candidate approaches. But you have not locked in vendors or architectures. These are funded explorations with decision gates at the end of Q2.

Quarter 5 and 6: Strategic themes only. 'We will apply AI to fraud detection in commercial lending.' Not 'We will deploy a fine-tuned LLM on transaction narratives using vendor X.' The theme stays stable. The implementation approach stays open.

Every quarter, you roll forward. Q3 bets become Q1 commitments with full detail. Q5 themes become Q3 bets with funded exploration budgets. New themes enter at Q5 and Q6 based on what you learned.

How the 6-Quarter Model Plays in Practice

I helped a mid-size bank implement this in 2024. Their previous approach was a 36-month roadmap that went stale every quarter but nobody wanted to formally revise because the governance process took 8 weeks.

With the rolling model, they run a quarterly strategy session. Two hours. The AI leadership team reviews what shipped, what learned, and what changed in the market. They promote directional bets to execution commitments. They retire themes that no longer make sense. They add new themes based on business demand or technology shifts.

In the first year, they retired three initiatives that would have consumed $800K in the old roadmap. Not because the ideas were bad, but because better approaches emerged that solved the same business problem at 40% of the cost. One fraud detection project originally scoped as a custom model build got replaced by a vendor API that did not exist when the theme was first added.

Their deployment count went from 2 production AI systems to 7 in twelve months. Not because they worked faster. Because they stopped defending outdated plans and started making decisions with current information.

The CFO loved it because every dollar had a shorter path to measurable outcome. No more 'trust us, it pays off in year three' conversations. Every commitment was 6 months or less to measurable value.

Making This Work With Your Governance Structure

The biggest objection I hear: 'Our board wants a long-term AI strategy.' Fine. Give them one. But separate the strategy narrative from the execution plan.

The strategy narrative is a 5-page document that says: here are the three to five business domains where AI creates competitive advantage for us. Here is our risk posture. Here is our talent model. Here is our vendor philosophy. That document changes maybe once a year.

The execution plan is the rolling 6-quarter model. It lives in a dashboard, not a PowerPoint. It shows what is committed, what is being explored, and what is on the horizon. Board members see the strategy narrative annually and the execution dashboard quarterly.

This separation solves the real problem. Boards want confidence that leadership has a direction. They do not actually want to approve specific model architectures 30 months in advance. Give them direction at the top and adaptability underneath.

One more structural piece: assign each quarter's commitments a single accountable executive. Not a committee. Not a working group. One name. When Q3 bets get promoted to Q1 commitments, somebody's name goes on the line. This is where most rolling plans fail. They stay flexible but never get specific about who owns what. Flexibility without accountability is just drift.

Actionable Takeaway