Strategy vs. Plan: A Foundational Distinction

Research leaders preparing to guide their institutions through an AI transformation often conflate two distinct activities: developing a strategy and developing a plan. This conflation is more than semantic. It produces fundamentally different documents with different functions, different processes for creation, and different criteria for success.

A strategy answers the questions of why and where: Why is AI research a priority for this institution, specifically, at this moment in time? Where will the institution invest: which AI domains, which research communities, which types of infrastructure and talent? A strategy is defined by choices: the decision not to pursue certain AI research areas is as important as the decision to pursue others. An institution that says it will pursue AI in all directions has no strategy, only aspiration.

A plan answers the questions of how and when: How will the strategic choices be executed: with what resources, what organizational structures, what implementation sequence? When will specific milestones be achieved, by whom, at what cost? A plan is a commitment to action; a strategy is a framework for decision-making. The plan should be derived from and anchored to the strategy, but the two documents serve different audiences and operate on different timescales. Strategy typically operates on a 3-to-5-year horizon; plans typically operate on a 12-to-18-month horizon and are revised annually.

The practical consequence of confusing strategy with plan is that institutions produce long, detailed action item lists that masquerade as strategies. These documents articulate what will be done but not why. They cannot guide decision-making when circumstances change, resources fall short, or unexpected opportunities emerge. A research leader confronting an unexpected AI partnership opportunity needs a strategy to evaluate whether the opportunity fits; an action item list provides no evaluative framework.

At the Level 5 (institutional leadership) level, developing a coherent AI research strategy is perhaps the single most consequential act a VPR or provost can perform for the institution's long-term competitiveness. The strategy becomes the basis for resource allocation, faculty recruitment, partnership development, and external communication. It takes on a life of its own: shaping thousands of individual decisions made by faculty, administrators, and students over the strategy period.

The Strategy Development Process: Five Phases

Developing a credible institutional AI research strategy requires a structured process that includes environmental scanning, internal capability auditing, scenario planning, strategy formulation, and stakeholder review. Skipping phases, particularly the environmental scan and scenario planning, consistently produces strategies that are either uninformed by external reality or too brittle to withstand unforeseen changes in the AI landscape.

Phase 1: Environmental Scan (4-6 weeks). The environmental scan examines four dimensions of the institution's external context. The technology landscape dimension tracks the current state and near-term trajectory of AI capabilities relevant to the institution's research strengths: which model architectures are becoming tractable, which compute paradigms are emerging, which datasets are becoming available, and which AI application domains are opening up. The competitive positioning dimension maps where peer institutions are investing in AI research, which faculty they are recruiting, which partnerships they are building, and which federal programs they are pursuing. The regulatory environment dimension monitors funding agency policies (data management requirements, research integrity standards), government AI executive orders, and evolving AI governance frameworks that will shape what is permissible and what is required. The funding opportunities dimension maps the current portfolio of federal AI research programs, their priorities, their funding levels, and their likely trajectory.

Phase 2: Internal Capability Audit (2-3 weeks). The internal audit documents the institution's existing AI research strengths, faculty interests, infrastructure, and partnerships. This phase draws on the AI Readiness Index assessment (covered in Lesson 1.3) and supplements it with qualitative interviews of department chairs and center directors about their AI research visions. The internal audit should produce an honest inventory of what the institution is already doing well and where it has real competitive advantage, not wishful thinking about aspirational capabilities.

Phase 3: Scenario Planning (2 weeks). Scenario planning is the most intellectually challenging phase and the most frequently omitted one. The team develops three plausible futures for the AI research landscape over the strategy period (3-5 years): a baseline scenario (AI capabilities continue advancing at roughly current rates, federal funding remains stable, regulatory environment remains manageable), an acceleration scenario (AI capabilities advance much faster than anticipated, compute costs drop dramatically, new application domains open rapidly, requiring much faster institutional adaptation), and a disruption scenario (regulatory restrictions tighten significantly, a major funding category is eliminated, or a competitor achieves a breakthrough that changes the competitive landscape). For each scenario, the team asks: what would this mean for our institution's AI research priorities? which of our proposed strategic moves would succeed in all three scenarios? which are contingent on only one scenario? This analysis surfaces the strategy's most critical assumptions and identifies strategic moves that are robust across multiple scenarios.

Phase 4: Strategy Formulation (cross-functional workshop, 2-day offsite). Strategy formulation happens in a structured 2-day offsite workshop that brings together a cross-functional leadership team: VPR, CIO, CFO or budget officer, three to five faculty leaders representing different disciplines, the graduate dean, and the head of research development. The workshop is facilitated (ideally by an experienced external facilitator) and follows a structured agenda that builds from environmental scan findings through internal capability assessment to scenario planning insights, then guides participants through a series of structured decisions about strategic focus areas, priority investments, and institutional positioning. The output of the workshop is a draft strategy framework with identified pillars, priority choices, and early-stage OKRs.

Phase 5: Stakeholder Review (4-6 weeks). The draft strategy framework must go through meaningful stakeholder review before it is finalized. This is not a rubber stamp process. It is a genuine opportunity for the faculty senate, graduate school, and relevant board committees to identify gaps, challenge assumptions, and propose modifications. Institutions that skip or rush the stakeholder review phase typically discover the consequences 12 to 18 months into implementation, when faculty who feel unheard become obstacles rather than advocates. The stakeholder review should have defined timelines (not open-ended) and a clear process for how feedback will be incorporated and how final decisions will be made.

Four Strategic Pillar Types: Building the Framework

The core of an institutional AI research strategy is a set of four to five strategic pillars that organize the institution's commitments across the key dimensions of AI research capability. While every institution's specific pillar content will differ based on its research strengths and strategic context, the four pillar types that consistently appear in effective institutional AI research strategies are: Research Focus, Talent Investment, Infrastructure, and Partnerships.

Research Focus Pillar. The Research Focus pillar answers: which AI research domains will this institution pursue, and which will it not pursue? The emphasis on 'and which will it not pursue' is deliberate. An institution that lists every major AI research area as a priority has made no choices, and choices are the essence of strategy. A Research Focus pillar for a health sciences-strong university might commit to: (1) AI for clinical decision support and precision medicine, (2) AI for biomedical image analysis and drug discovery, and (3) trustworthy AI systems with applications in health contexts. These three choices represent a coherent cluster of related AI research domains that build on the institution's existing strength in biomedical research. They also implicitly decline certain AI domains, natural language processing for general-purpose applications, AI for robotics and autonomous systems, and AI for financial technology, where the institution has no distinctive advantage.

Talent Investment Pillar. The Talent Investment pillar covers: who will the institution hire, how will existing faculty be supported in developing AI research capacity, and how will AI talent be retained? Typical talent pillar elements include: a 3-year faculty hiring plan (number of cluster hires, target departments, seniority levels), a graduate training strategy (new PhD concentrations in AI methods, master's programs with AI specialization, interdisciplinary fellowships), and a professional development program for existing faculty who want to incorporate AI methods into their research programs. The talent pillar should explicitly address retention as well as recruitment: the talent market for AI-skilled faculty is extraordinarily competitive, and an institution that recruits AI faculty without attending to competitive compensation, research support, and collaborative environment is likely to lose them to both peer institutions and industry within 3-5 years.

Infrastructure Pillar. The Infrastructure pillar defines the institution's commitment to compute, data, and software infrastructure required to enable the Research Focus areas. It should specify: the target compute capacity and the build/buy/share strategy for achieving it, the research data management approach, the software and tooling stack the institution will support, and the governance structure for managing shared infrastructure. The Infrastructure pillar is the most capital-intensive element of most AI research strategies and is typically where the tension between strategic ambition and fiscal constraint is sharpest. The pillar should include both a fully funded scenario and a constrained scenario to guide decision-making if available resources fall short of the full vision.

Partnerships Pillar. The Partnerships pillar defines the institution's approach to external collaboration: which partner types will be prioritized (industry, federal, international, consortial), what the target portfolio mix looks like, and what the partnership development and management infrastructure will be. The Partnerships pillar should be directly linked to the Research Focus pillar: the partnership targets should be those most likely to extend the institution's capabilities in its chosen AI research focus areas, not a generic 'we will pursue partnerships' commitment.

The 3-5 Year AI Research Roadmap: Foundation, Scaling, Leadership

Translating strategy into a roadmap makes the strategy actionable. The roadmap organizes the strategy's initiatives into a time-sequenced set of phases that build institutional capability progressively. Most effective institutional AI research roadmaps are organized into three phases: Foundation, Scaling, and Leadership.

Foundation Phase (0-18 months). The Foundation phase focuses on establishing the core capabilities required to enable subsequent phases. Foundation investments are not necessarily visible externally. They are the unglamorous enabling conditions that make more visible accomplishments possible. Typical Foundation phase activities include: completing the AI Readiness Assessment and establishing the baseline scorecard; hiring the first cohort of enabling talent (research software engineers, data scientists, ML engineers); standing up pilot AI infrastructure (often beginning with a cloud-based research computing environment while on-premise procurement is planned); adopting AI-specific governance policies (AI use policy, data governance framework, IRB guidance); and launching one to three pilot AI research projects that demonstrate the institution's emerging capabilities and generate early publications. The Foundation phase should end with a review against baseline readiness metrics to confirm that the enabling conditions for the Scaling phase are in place.

Scaling Phase (18-36 months). The Scaling phase expands the initiatives piloted in the Foundation phase to broader institutional scope and higher resource levels. Compute infrastructure built in the Foundation phase is expanded to serve a broader researcher community. The hiring plan continues, with second-cohort hires building depth in priority AI research areas. External grant submissions, informed by the pilot projects and relationships from the Foundation phase, ramp up significantly. Partnership agreements signed or initiated in the Foundation phase move into active execution, generating research outputs that build external reputation. Graduate AI research programs that were proposed in the Foundation phase are now enrolling students. The Scaling phase is when the institution begins to experience the compounding effects of strategic AI investment: publications generate citations and collaborator interest, which generate grant proposals, which fund more research, which generates more publications.

Leadership Phase (36-60 months). The Leadership phase is when the institution becomes a recognized leader in its chosen AI research focus areas. Leadership does not mean leading every AI research area. It means being recognized as a first-mover and thought leader in the specific domains the strategy committed to in the Research Focus pillar. Leadership indicators include: being named as a preferred partner by both industry and federal program managers, attracting top AI-adjacent faculty as primary choices (rather than accepting hires that didn't go to R1 competitors), hosting or co-organizing major conferences in the focus areas, and generating research that influences standards, policy, or commercial practice in the focus domains.

Governance for the Strategy: Ownership, Review Cycles, and Board Reporting

A strategy document without governance infrastructure is a paper document. Institutional AI research strategies require explicit governance arrangements that specify who owns the strategy, how it is reviewed and updated, and how progress is reported to the board and other accountability bodies.

Strategy Ownership. The institutional AI research strategy should have a named owner: typically the VPR, though at some institutions with a dedicated chief AI officer or chief research officer, that role may be more appropriate. The owner is responsible for: convening the annual strategy review, maintaining the action plan, escalating emerging issues to the provost or board, and representing the strategy in external stakeholder communications. Strategy ownership should not be diffused across a committee, committee ownership means no one person is ultimately responsible, which allows the strategy to drift.

Review Cycles. The strategy should be reviewed on two cycles: a quarterly operational review (focused on action plan execution, OKR progress, and emerging issues requiring tactical response) and an annual strategic review (focused on whether the strategic pillars remain appropriate given changes in the technology landscape, competitive positioning, and available resources). The annual strategic review is also when major changes to the strategy, such as adding a new Research Focus area in response to an emerging AI domain, or adjusting the Infrastructure pillar in response to compute cost changes, should be formally made and documented.

Board Reporting. The board of trustees should receive an annual strategy update covering: overall progress against the roadmap milestones, key achievements of the past year (publications, partnerships, grants, hires), resource investment and return, emerging risks and the response plan, and the year-ahead priorities. Board members at research universities increasingly have expectations for AI strategy literacy; the strategy report should include enough context about the AI research landscape to allow board members to exercise meaningful oversight rather than simply ratifying staff recommendations.

Strategy Adaptation Triggers. Certain events should trigger a formal review of whether the strategy needs to be adapted: a major technology discontinuity (a new model architecture or compute paradigm that changes the landscape), a significant competitor move (peer institution announces a $100M AI gift or major AI institute award), a major funding change (a key federal AI program is eliminated or radically reprioritized), or a significant internal event (the departure of a key faculty leader or the completion of a major infrastructure build that changes the institution's capability profile). Adaptation triggers should be documented in the strategy document itself, so that the leadership team has a shared understanding of when revisiting the strategy is appropriate rather than treating every change as a potential reason to abandon strategic commitments.

Communication Strategy: Differentiated by Audience

An institutional AI research strategy is only as effective as its communication. A strategy that remains in a document and is not understood or embraced by the people responsible for executing it will not change institutional behavior. Communication of the strategy must be differentiated by audience, because different stakeholders need different aspects of the strategy and receive it through different channels.

Provost and Senior Leadership. The provost and senior leadership team need the strategic framework, the pillar structure, the roadmap phases, the resource requirements, and the board-ready narrative. They need to understand the strategy well enough to represent it in external conversations, in budget negotiations with the CFO, and in governance discussions with the board. Communication with this group is typically through direct briefings and a written strategy brief (4-6 pages) supplemented by an executive summary for board use.

Faculty. Faculty communication requires a fundamentally different approach than executive communication. Faculty are simultaneously the primary beneficiaries of the strategy and its primary executors. They conduct the research, recruit the graduate students, submit the grants, and produce the publications that the strategy is designed to enable. They are also a highly skeptical audience that is professionally trained to identify logical gaps and untested assumptions. Faculty communication should emphasize the research opportunities the strategy creates: funding priorities aligned with faculty interests, infrastructure investments that enable new research directions, talent support that reduces administrative burden. Faculty information sessions, one per college or school, not one campus-wide meeting, allow for discipline-specific questions and create a sense of direct engagement rather than top-down mandate.

Graduate Students. Graduate students need to understand how the AI research strategy affects their own training, research opportunities, and career preparation. The graduate student communication should emphasize: new AI research projects they can join, new interdisciplinary AI training programs being developed, industry partnership opportunities that create internship and placement pathways, and the institution's commitment to computing resources that support dissertation-level AI research.

Board of Trustees. Board communication requires translating the academic AI research strategy into the language of institutional mission, competitive positioning, and fiduciary responsibility. Board members are not AI researchers. They need to understand the strategic context (why AI matters for research universities), the institutional position (where this institution stands relative to peers), the strategic response (what the institution is doing about it and why it is the right response), and the resource requirements and expected returns. The board presentation should be no longer than 20 minutes plus Q&A and should be story-driven rather than data-dense.

Strategy Execution Infrastructure: OKRs and Quarterly Business Reviews

The gap between strategy and execution is the most common strategic failure mode in academic institutions. An impressive strategy document with beautiful slides produces no research impact if the institution does not put in place the execution infrastructure to translate strategic commitments into day-to-day action. Two critical execution infrastructure elements are an OKR framework and a quarterly business review process.

OKRs Aligned to Strategic Pillars. OKRs, Objectives and Key Results, are a goal-setting methodology that has proven effective in complex organizations. Each strategic pillar should have a corresponding OKR set that specifies: the Objective (a qualitative statement of what the pillar aims to achieve), and two to five Key Results (specific, measurable outcomes that indicate progress toward the Objective). For the Research Focus pillar, an Objective might be: 'Establish the institution as a nationally recognized leader in AI for precision medicine research.' Key Results for this objective might include: five publications in Nature Medicine, NEJM, or equivalent top journals featuring AI methods (by month 18), two NIH R01 awards specifically using AI approaches in precision medicine (by month 24), and one joint research agreement with a pharmaceutical company or health system for AI-powered clinical research (by month 30). OKRs provide the measurement framework that makes strategy accountability possible and gives the quarterly review process something concrete to evaluate.

Quarterly Business Reviews (QBRs). A Quarterly Business Review is a 90-to-120-minute leadership meeting, held every quarter, that reviews progress against the OKRs for each strategic pillar. The QBR format is structured: each pillar owner (typically a faculty or administrative leader with named ownership of a specific pillar) presents a 10-to-15-minute update covering key results status (on track / at risk / behind), the most significant action taken in the past quarter, the most significant obstacle, and the top priorities for the next quarter. The VPR (or whoever owns the strategy) facilitates the overall review and makes real-time decisions about resource reallocation or escalation for any issues that cannot be resolved at the pillar owner level. QBRs prevent the common failure mode of strategy drift, where execution gradually diverges from strategic intent as day-to-day pressures push organizations toward familiar activities rather than strategic priorities.

Strategy in Practice: Institutional Examples

Understanding AI research strategy at the institutional level is enriched by examining how different types of institutions have approached the challenge. The following case examples illustrate how strategy principles translate into specific institutional choices.

MIT: Deep Technical Leadership with Deliberate Boundary-Setting. MIT's AI research strategy, operationalized most visibly through the MIT Stephen A. Schwarzman College of Computing (launched with a $1B commitment in 2018 and reshaping the entire academic structure), reflects a clear strategic choice to pursue deep technical AI leadership across multiple fundamental domains while simultaneously integrating AI into every academic discipline. The MIT strategy made explicit choices about what the college would and would not do, it is primarily a research and education enterprise, not a commercialization engine, and structured faculty appointments to enable genuine cross-disciplinary collaboration rather than siloed departmental research. The strategy's coherence derives from the alignment of its structural innovation (the College of Computing itself), its resource commitment (new faculty lines, new building, dedicated computing infrastructure), and its intellectual vision (computing and AI as a second language for all MIT students).

Carnegie Mellon University: Portfolio Breadth with Ecosystem Depth. CMU's AI research strategy leverages the institution's extraordinary depth in both computer science and adjacent fields (robotics, human-computer interaction, language technologies) to pursue a broad AI research portfolio anchored by the School of Computer Science and extending across nine degree-granting units. CMU's strategy creates strategic value through ecosystem effects, the concentration of AI talent creates a research culture and collaboration density that amplifies individual researchers' productivity. The CMU strategy is less defined by explicit pillar choices and more defined by talent concentration and ecosystem investment.

Smaller Institutions: Niche Leadership Strategy. Smaller research universities cannot pursue the breadth strategies of MIT or CMU, but they can win in specific niches. A regional university with strong medical research programs and a health system partner can become a recognized leader in clinical AI applications, AI for imaging interpretation, AI for patient risk stratification, AI-assisted diagnostics, within its specific patient population and institutional context. The key strategic insight for smaller institutions is that national leadership in a narrow domain is more achievable and more valuable than mediocrity across a broad domain. A niche leadership strategy requires even more disciplined 'not doing' choices than a broad strategy, because the resource constraints are tighter and the temptation to expand scope is constant.

Strategy Adaptation: Staying Relevant in a Rapidly Evolving Landscape

AI research strategy in 2026 faces an unusual challenge: the pace of technological change in AI is faster than most strategic planning cycles. A strategy developed in 2024 may need significant adaptation by 2026 based on the emergence of new model architectures, new compute paradigms, new regulatory requirements, or new competitive dynamics that were not foreseeable at the time of strategy formulation.

Effective strategy adaptation requires distinguishing between three types of change signals: noise (short-term fluctuations that don't warrant strategic response), trends (consistent directional changes that should inform the next annual strategy review), and discontinuities (step-change events that warrant immediate strategy review and possible pivoting). The emergence of a new open-source model family from a well-funded lab is probably a trend worth monitoring rather than an immediate strategy change driver. A major new federal funding program specifically targeting the institution's chosen AI research focus areas is a discontinuity worth an immediate strategy review to assess how best to position for the opportunity.

Institutions that update their AI strategy too frequently, in response to every AI news cycle, signal to faculty, partners, and funders that they lack strategic conviction and cannot maintain sustained investment in their chosen priorities. Institutions that never update their strategy, treating the initial strategy as a permanent commitment regardless of how the landscape changes, risk irrelevance as their strategic assumptions become obsolete. The right cadence, informed by the adaptation triggers specified in the strategy governance section, is an annual review cycle with trigger-based interim reviews for discontinuities.

Finally, strategy adaptation must be distinguished from strategy abandonment. When the landscape shifts, the appropriate response is usually to adjust emphasis within the existing strategic pillars rather than to abandon the pillars entirely. An institution that committed to AI for precision medicine as a Research Focus pillar should respond to a new federal precision medicine data initiative not by abandoning the pillar but by repositioning the pillar to take advantage of the new initiative. Strategy adaptation is about maintaining the coherence and commitment that make strategy valuable while updating specific implementation approaches in response to changed circumstances.