How to Sequence an AI Investment Plan With Your Board Across 3 Years

Why a Three-Year Horizon

The three-year horizon is not arbitrary. Shorter horizons (twelve months) do not allow the foundation work to mature into portfolio leverage; longer horizons (five years) face too much technology and regulatory uncertainty to commit specific investment levels with credibility. Three years is the window in which the foundation can be built, scaled, and producing portfolio leverage, while the technology and regulatory environment can be reasonably anticipated.

The three-year structure also aligns with how most boards plan capital allocation. Annual reviews are too frequent for capability builds that take eighteen to twenty-four months to mature; five-year plans are too long to be ratified with conviction. A three-year plan with explicit annual checkpoints fits the board governance cycle and produces decisions the board can stand behind.

The sequencing across the three years is the discipline that distinguishes credible plans from aspirational ones. Each year must accomplish something specific, the accomplishments must build on each other, and the milestones must be measurable enough to determine whether the next year’s investment should accelerate, hold, or pull back. The BCG analysis of AI value capture emphasizes this point explicitly: organizations that capture meaningful AI value sequence their investments deliberately, and organizations that fail to capture value often invest aggressively in all three years simultaneously, producing diffuse activity without compounding capability.

Year One: Foundation and Proof Points

Year One is the foundation year. The objective is to build the capability that subsequent years will leverage, not to deliver portfolio-wide impact. Boards that expect Year One to produce material P&L impact typically push the program into overreach; boards that understand Year One as foundation produce better outcomes.

What Year One should accomplish:

• Governance infrastructure. An AI governance committee chartered with defined decision rights, an AI use case inventory, a tier classification SOP, and the policy artifacts that the broader portfolio will sit on. This is unglamorous but irreplaceable foundation.
• Three to five proof-point use cases. Use cases selected through the prioritization matrix discipline, with explicit success criteria, defined budgets, and end-of-year evaluation gates. The proof points are not the strategic AI investments; they are the disciplined demonstrations that the foundation can support real use cases.
• Data infrastructure investments. The data preparation, governance, and integration work that the proof points reveal as binding constraints. Most organizations underinvest in this layer in Year One; the foundation does not hold without it.
• Workforce competency baseline. Initial training and competency development across the functions that will participate in AI use cases, including QA, IT, and use case owners. This is the start of the workforce build, not the completion.
• Vendor and partnership relationships. The contracts, partnerships, and external relationships that will support Year Two scaling. Negotiating these in Year One when the program has less urgency produces materially better terms than negotiating in Year Two under deadline pressure.

The financial structure for Year One typically allocates 60-70% to foundation work (governance, data, infrastructure, partnerships) and 30-40% to the proof-point use cases themselves. Programs that invert this ratio — heavy investment in use cases with minimal foundation work — consistently struggle to scale in Year Two because the foundation is not present to scale onto.

Year Two: Scaling and Operating Model

Year Two is the scaling year. The Year One proof points either succeed or do not; either way, the Year Two investment is informed by what the proof points revealed. The objective for Year Two is to scale the demonstrated capability across the broader portfolio while stabilizing the operating model that makes scaling sustainable.

What Year Two should accomplish:

• Scaled portfolio. Six to twelve active use cases at meaningful depth, drawn from the prioritization matrix and informed by Year One’s empirical learning about which categories of use cases the organization can actually deliver.
• Operating model stabilization. The cross-functional working patterns, decision rights, escalation paths, and review cadences that make the portfolio operationally sustainable. This is the work that turns AI from a project portfolio into an operating discipline.
• QA and validation capacity build. The deeper QA capability development, validation methodology refinement, and inspection-readiness work that the Year One proof points typically reveal as gaps. As discussed in the broader McKinsey research on AI operating models, this capability is what distinguishes organizations that scale from organizations that get stuck at proof-of-concept.
• Performance monitoring and lifecycle management infrastructure. The monitoring, telemetry, and lifecycle management work that production AI requires. Year One typically defers this; Year Two cannot.
• External engagement and credibility. Engagement with industry working groups, regulatory engagement on Year Two use cases that require it, and the credibility-building work that supports Year Three’s more ambitious portfolio. As the McKinsey State of AI research documents, organizations capturing significant AI value almost always engage externally as part of their scaling, not after it.

The Year Two financial structure typically shifts toward direct use case investment (40-50%) with continued foundation investment (30-40%) and increasing investment in operating model capability (15-25%). The ratio shift reflects the maturation: Year Two is when the foundation actually starts to be leveraged.

Year Three: Portfolio Leverage

Year Three is the leverage year. The foundation built in Year One and the scaling achieved in Year Two should produce a portfolio that delivers compounding value across multiple use cases sharing common infrastructure, capability, and governance. The objective is to demonstrate the portfolio leverage that justified the cumulative investment.

What Year Three should accomplish:

• Full portfolio at scale. Twelve to twenty active use cases across the strategic priorities, delivered through the operating model stabilized in Year Two and sharing the foundation built in Year One.
• Measurable enterprise impact. The P&L, operational, and strategic impact metrics that the three-year plan promised at the outset. By Year Three, the cumulative impact should be visible at the board reporting level, not just at the program level.
• Capability that is durable beyond the original program team. The AI capability should be embedded in the organization rather than depending on a specific program team. Boards that ask whether the capability would survive the departure of the original program lead are asking the right question.
• Forward portfolio that is not dependent on continued exceptional investment. Year Three should produce a portfolio that continues to deliver value without requiring the investment intensity of Years One and Two. The foundation should be substantially complete; the operating model should be stable; the capability should be embedded.
• Next-horizon planning. By the end of Year Three, the program should be positioned to articulate the next horizon — whether that is continued scaling, strategic AI investments at the next level of ambition, or selective deepening in specific high-value domains. The forward direction should be a deliberate choice, not a reactive one.

The Year Three financial structure typically shifts to majority use case investment (50-60%) with maintenance of the foundation and operating model investment (30-40%) and increasing investment in next-horizon capability (10-20%). The structure reflects portfolio maturity: the foundation is paying off, and the marginal investment is in extending the portfolio rather than building new foundation.

The Financial Structure That Holds Up

The financial structure of the three-year plan needs to satisfy three audiences: the operating team that has to execute against it, the finance function that has to ratify it, and the board that has to commit capital to it. The structure that has worked across our client engagements has several recognizable features.

The cumulative investment over the three years is meaningful — typically 1.5-3% of relevant operating expense for the functions in scope, though this varies significantly by industry and starting state. Boards that focus on the absolute dollar figure without the capability state context often pull back at the wrong moment; boards that understand the year-over-year capability progression typically commit with appropriate conviction.

Milestones That Determine Acceleration or Pull-Back

The three-year plan should include explicit milestones at the end of each year that determine whether the next year’s investment should accelerate, hold, or pull back. Without these milestones, the plan becomes a commitment that runs regardless of performance; with them, the plan becomes a structured commitment that adjusts based on demonstrated learning.

End of Year One milestones:

• Governance infrastructure is operational and producing decisions, not just procedural compliance
• At least 60% of the proof-point use cases are delivering measurable value at the use case level
• Data infrastructure investment is producing accessible, governed data for the prioritized use cases
• Workforce competency development is underway with measurable progress
• Vendor and partnership relationships are operational, not just contractual

End of Year Two milestones:

• The scaled portfolio is delivering measurable value at the use case level across the majority of the active use cases
• The operating model is producing decisions, deployments, and lifecycle management without requiring exceptional intervention
• QA and validation capacity has matured enough to support the Year Three portfolio scale
• Performance monitoring is operating across production use cases with defined response procedures
• External engagement is producing concrete benefit for the program, not just nominal participation

End of Year Three milestones:

• Enterprise-level impact is visible at the board reporting cadence
• Capability is embedded enough to survive the departure of key program personnel
• Forward portfolio does not require Year One-level investment intensity to sustain
• Next-horizon planning is deliberate and well-grounded in three years of empirical learning

Each milestone is a check on whether the program is delivering against the foundation it has built. Programs that meet the milestones earn the conviction to accelerate; programs that miss them produce the data needed to pull back or restructure. The discipline of the explicit milestones is what makes the three-year plan adjustable rather than rigid.

Conducting the Board Conversation

The board conversation around a three-year AI investment plan has a recognizable structure when it works well. The pattern that has been most effective across client engagements.

The plan is presented as a sequenced capability build, not as a list of projects. The framing matters: a project list invites scrutiny of individual line items; a capability build invites the strategic question of whether the trajectory is right. Boards that adopt the capability framing make better decisions than boards that drill into individual use case detail.

The plan articulates explicit hypotheses about what will be true at each year-end. These hypotheses are testable; they form the basis for the milestone review at each year-end. Boards appreciate the explicit hypothesis discipline because it makes the program accountable in a way that aspirational statements do not.

The plan articulates the financial structure with the foundation/use case/operating model split visible. The split is important because it makes the trade-offs explicit and prevents the board from focusing only on the direct use case investment.

The plan includes the pull-back scenarios. Boards that have seen the program articulate what it would do if Year One milestones are not met — pull back to a smaller portfolio, restructure the foundation work, replace specific elements — commit more confidently than boards that have only seen the success scenario. The pull-back articulation signals discipline, not lack of conviction.

The plan is presented by the cross-functional leadership team, not just by the AI program lead. The cross-functional presentation signals that the plan is owned across the organization, not by a single program. This is particularly important for pharma and biotech clients, where quality, regulatory, IT, and the business functions all have material stakes in the plan’s execution.

As noted in Harvard Business Review’s analysis of generative AI risk management, boards that understand AI investment as a capability build rather than a series of discrete projects consistently make better governance decisions. The three-year sequencing structure is one of the most direct mechanisms for producing that framing.

How the plan adapts to different organizational starting states

An important calibration: the three-year plan structure is consistent across organizational starting states, but the specifics within each year vary materially. Organizations starting from minimal AI capability spend Year One almost entirely on foundation, with proof points kept deliberately modest. Organizations starting with established AI capability spend Year One refining the foundation and conducting more ambitious proof points. The structure is the same; the content within each year reflects the starting state.

For pharma and biotech clients specifically, the starting state often differs across the organization. R&D may have substantial AI capability while manufacturing has minimal capability. The three-year plan can be structured at the function level or at the enterprise level, but the relationship between function-level plans and the enterprise plan should be made explicit. Function-level plans that are not aligned to an enterprise plan produce duplication and gaps; enterprise plans that do not respect function-level starting states produce unrealistic expectations.

The role of external benchmarks in board conversations

Boards reviewing the three-year plan will often want to understand how the proposed investment compares to peer organizations. External benchmarks — from McKinsey, BCG, Deloitte, and similar sources — provide useful context but should not be the primary basis for the plan. The benchmarks describe what peer organizations are doing on average; they do not describe what is right for the specific organization given its specific starting state, strategic priorities, and operational constraints.

The right use of external benchmarks in the board conversation is as triangulation: confirming that the proposed investment is within the range of peer practice, rather than as the primary justification. Plans that are anchored entirely in external benchmarks lack the specificity that distinguishes a defensible plan from a generic plan; plans that ignore external benchmarks entirely produce uncomfortable conversations when board members ask the natural comparison question. The discipline is to use the benchmarks as one input among several, with the organization-specific analysis carrying the primary weight.

Communicating progress between board reviews

The board conversation does not end at plan ratification; it continues through the three years via the communication cadence the program establishes. The cadence we recommend with clients includes quarterly board updates that report against the explicit milestones, mid-year deeper reviews at the end of Year One and Year Two, and the formal year-end milestone reviews that determine the next year’s structure.

The quarterly updates should be short — typically two to three pages — and structured around the milestone progress, the financial actuals against plan, and the risks the program is managing. The mid-year deeper reviews should include richer narrative on the operational learning, the cross-functional dynamics, and the strategic implications of the year’s experience. The year-end milestone reviews should be the most substantive conversations, supporting the next year’s investment decisions with the full year of empirical evidence.

Programs that maintain this communication cadence produce board conversations that build conviction over time. Programs that go silent between annual reviews produce board conversations that have to rebuild conviction at each annual review, and the absence of progressive engagement makes the year-over-year decisions more fragile than they need to be.

References & Sources