National AI Strategy Sprint

Format: Hybrid (curated readings + live discussion + async exercises) · Duration: 5 weeks + optional Week 2.5 deep dive · Target audience: Government officials, policymakers, analysts, strategists, educators, civil society, industry policy teams, and AI enthusiasts — Register for the Sprint here.

Why this program, and why now (2026)

Artificial intelligence has moved from the lab to the core of national power. It now shapes economic growth, national security, the quality of public services, and where a country sits in the alliances that will define the next two decades. This is no longer a technology story with a policy footnote. It is a statecraft story with a technology engine.

Governments that act with clarity and speed are writing the rules and capturing the value. Those that hesitate become rule-takers instead of rule-makers — importing standards, tools, and dependencies designed around someone else's constraints. A National AI Strategy is simply how a country decides, out loud and on the record, what it wants from AI and how it intends to get there responsibly.

Most national AI strategies fail in three predictable ways:

• Too vague. "We will be a leader in AI." A slogan is not a strategy. It names an ambition without naming a single choice, trade-off, or owner.
• Too narrow. A list of R&D grants and ethics principles with no compute reality, no institutional machinery, no labor plan, and no metrics. The interesting 80% of the problem is missing.
• Too copy-paste. An imported blueprint that assumes reliable power, cheap compute, deep talent pools, and fiscal room a country does not have — and quietly ignores the geopolitics of who owns the models, chips, and cloud.

CentPol's National AI Strategy Sprint is built to close that gap. Over five weeks you will learn the nine canonical pillars that appear again and again in the strongest strategies worldwide, practice applying them until the framework is second nature, and assemble a concise, context-aware strategy memo you could put in front of a real decision-maker.

The 2026 moment genuinely is different, and the difference favors newcomers:

• Regional coordination is maturing. The African Union and ASEAN have moved from principles to operational guidance, turning regional alignment into leverage rather than a constraint.
• The cost curve has bent. Model-efficiency breakthroughs and smaller open-weight models have lowered the price of adoption and experimentation, narrowing the gap between well-resourced and resource-constrained states.
• Safety capacity is spreading. AI safety institutes and evaluation capability are no longer a Western monopoly; the know-how is becoming a shared, buildable asset.

Together these create real room for African and Global South countries to shape outcomes on their own terms — to design strategy for their context rather than implement someone else's.

What this sprint delivers

By the end of the program, you will be able to:

• Decode any national AI strategy into its core building blocks, and spot the missing pieces others miss.
• Design an implementable strategy that accounts for compute, data, institutions, labor, and geopolitics — not just R&D and ethics.
• Produce decision-ready outputs: a two-page strategy memo and a short oral pitch built to survive a senior leader's attention span.

No coding background is required. This is built for people who work with technology policy and strategy, not necessarily inside engineering teams.

Who this program is for

• Government officials and advisers working on AI, digital, innovation, industrial, economic, or security policy
• Think tanks, NGOs, and international organizations supporting national strategy and implementation
• Industry associations, telecom and compute-ecosystem stakeholders, and corporate technology-policy teams
• Academics and educators supporting public policy and technology governance
• Practitioners across Africa, Southeast Asia, Latin America, the Middle East, South Asia, and other Global South contexts
• Voices working in AI safety, AI ethics, and AI policy and governance

You do not need to arrive with a country in mind, though it helps. You can also work a fictional-but-realistic case, or the context of an institution you know well.

Program format and time commitment

• Weekly time: ~2–3 hours of reading + ~1 hour live session + ~1 hour independent writing/exercise
• Live sessions: 60 minutes weekly (Week 2.5 and Week 5 run 90 minutes)
• Cohort mode: structured discussion + applied exercises + peer review
• Final deliverables:
• A two-page National AI Strategy memo (context-aware, implementation-focused)
• A three-minute oral pitch (no slides required)

The load is deliberately light on volume and heavy on judgment. You will read curated excerpts, not whole reports, and spend most of your effort applying what you read to a real decision.

The CentPol 9-Pillar Framework

Across effective strategies — resource-rich and resource-constrained alike — a common architecture appears again and again. CentPol organizes it into nine pillars. Treat this less as a table of contents and more as a diagnostic instrument: a set of nine questions you can run over any strategy (yours or someone else's) to find where it is strong, where it is hollow, and where it is silent.

For each pillar, hold four things in mind — the core question it answers, what good looks like, the most common failure, and one diagnostic question you can ask on the spot.

1. Vision & Narrative Core question: What future is the country aiming for, and why now? Good looks like: a specific, contestable claim about the future that a citizen and a finance minister can both understand. Common failure: aspiration as strategy ("become a leader") with no choice inside it. Ask: If I deleted the country's name, could this vision belong to any country on earth? If yes, it is a slogan.

2. Compute & Digital Infrastructure Core question: Where will the compute, connectivity, and energy come from — and on whose terms? Good looks like: a realistic path to access (own, rent, or partner) that names the sovereignty-versus-access trade-off out loud. Common failure: treating compute as line-item IT spend rather than strategic infrastructure, and ignoring the power grid entirely. Ask: If demand for compute tripled next year, what exactly happens — and who do we depend on?

3. Data Governance & Public Data Assets Core question: What data does the state hold, how good is it, and can it be shared safely and usefully? Good looks like: data treated as an asset with quality, interoperability, privacy, and trusted-sharing arrangements — plus a sovereignty position. Common failure: conflating "data" with "compute," or writing a privacy law with no plan to make public data usable. Ask: Name one high-value public dataset and the exact rule that governs who may use it. Silence here is the answer.

4. Talent, Skills & R&D Ecosystem Core question: Who will build, buy, govern, and use AI — across the whole workforce, not just researchers? Good looks like: a pipeline that spans elite research, practitioner skills, broad AI literacy, and a serious answer to brain drain. Common failure: funding a few PhD scholarships and calling it a talent strategy while the best people emigrate. Ask: What keeps a country's best AI graduate from taking the first offer abroad?

5. Priority Sectors & Public Sector Transformation Core question: Where do we apply AI first — and how does government itself modernize? Good looks like: two or three named sectors with specific first use cases, plus a plan to reform the state's own operations. Common failure: "AI in health, education, and agriculture" with no use case, owner, or sequence — everywhere means nowhere. Ask: Name the single Year-1 use case whose success would prove the whole strategy is working.

6. Governance, Ethics, Rights & Safety Core question: What rules, standards, oversight, and safety capacity keep this trustworthy? Good looks like: buildable machinery — audits, evaluations, incident response, redress — scaled to budget, not just a list of values. Common failure: a page of principles no institution is funded or mandated to enforce. Ask: When a deployed system causes harm, who finds out, who is accountable, and what happens next?

7. Institutions, Coordination & Implementation Machinery Core question: Which offices, agencies, and councils actually run the strategy — and who has the power to make ministries move? Good looks like: a named owner with authority, a real delivery mechanism, and a "minimum viable institution" you can stand up now. Common failure: a coordinating body with a mandate to "coordinate" but no budget, staff, or power to say no. Ask: Who can stop a ministry from ignoring the strategy — and what is that person's title?

8. International Engagement & Geopolitics Core question: Where does the country stand on alliances, standards, supply chains, and multi-alignment? Good looks like: a deliberate posture — regional alignment, vendor diversification, sovereignty red lines, and standards participation. Common failure: passive alignment by default, importing whatever the dominant vendor or bloc offers. Ask: What is the one thing we will not concede on, regardless of who is offering the deal?

9. Monitoring, Metrics & Review Core question: How will we know if this is working, and how will we correct course? Good looks like: three to five behavior-driving indicators, named owners, review cadence, and explicit pivot points. Common failure: vanity metrics (number of strategies published, workshops held) that measure activity, not outcomes. Ask: Which single number, if it moved the wrong way, would force us to change the plan?

This sprint uses the nine-pillar map every week. You will not just learn the framework — you will apply it until it becomes a reusable instrument you carry into your own work.

How the sprint is arranged

The five weeks are not five topics. They are one cumulative build. Each week hands you the same nine-pillar diagnostic, points it at a harder layer of the problem, and asks you to produce a fragment of your final memo. By Week 5 you are not starting a memo — you are assembling one you have been writing all along.

• Week 1 — The landscape. Vision and the "why now," with Africa and the Global South at the center. You emit: your vision and urgency narrative.
• Week 2 — The architecture. The nine-pillar map, learned by comparing real strategies. You emit: a pillar map of your context.
• Week 2.5 (optional) — The constraints. Leapfrogging and honest budgets under real limits. You emit: your binding-constraint list.
• Week 3 — The engines. Compute, data, and institutions, designed to actually run. You emit: your institutional wiring and first-100-days.
• Week 4 — The application. Sectors, society, labor, and international stance. You emit: your priority sectors and posture.
• Week 5 — The synthesis. Metrics, iteration, and the decision-ready memo. You emit: the finished memo and pitch.

WEEK-BY-WEEK PROGRAM (2026)

Week 1 — The landscape: vision and the "why now"

Goal: Understand the global AI landscape and write a crisp, context-grounded vision and urgency narrative.

The idea

Every serious strategy opens with an argument, not a mood. The argument has two halves: where we actually stand (grounded in evidence, not vibes) and why acting now beats acting later (grounded in a specific risk of waiting). Most vision statements fail because they skip the evidence and reach straight for the aspiration — "we will be a leader" — which is unfalsifiable and therefore unpersuasive.

The discipline this week is to build a vision that survives a skeptic. A finance minister will ask "compared to what, and why this year?" Your narrative has to answer both. That means anchoring urgency in real numbers (investment flows, capability trends, adoption rates) and in your own context's exposure — what you lose by a year of delay that you cannot recover later. The 2026 framing matters here: falling model costs and maturing regional coordination mean the window for agency is open now in a way it was not three years ago, and may not be in three more.

For Africa and the Global South, the sharpest version of "why now" is about rule-making versus rule-taking. Standards, data-sharing norms, and compute dependencies are being set this decade. A country that shows up with a coherent position helps write them; a country that waits inherits them.

Anchor readings

• African Union: Continental AI Strategy (Executive Summary + structure)
• What it is: The AU's continent-wide AI strategy — a shared framing, priority areas, and coordination logic meant to guide national strategies and regional cooperation.
• Why it matters: It is the closest thing to a reference architecture for African AI agency (sovereignty, inclusion, capacity building), and it treats regional alignment as a strategic asset rather than a constraint.
• Contribution: Anchors us in Africa-first premises — what "good" looks like when infrastructure gaps, data-extraction risk, and regional bargaining power are front and center.
• Stanford HAI: AI Index Report (latest edition, Executive Summary)
• What it is: A short, visual overview of global trends in AI capabilities, investment, regulation, and public opinion.
• Why it matters: It gives you real numbers to ground your narrative — who is ahead, where investment is flowing, and what is changing fastest.
• Contribution: Lets you justify urgency and opportunity in your own context: "here is where we stand, and here is what we risk if we wait."
• OECD: AI Principles (summary)
• What it is: A concise set of internationally agreed principles for trustworthy AI (fairness, robustness, transparency, accountability, human-centered values).
• Why it matters: Many national strategies explicitly reference or align with these principles; they form a widely recognized normative baseline.
• Contribution: Gives you language and concepts you can safely reuse in any democratic or multilateral context.
• Carnegie (Africa program): Africa's AI governance landscape (policy-to-practice framing)
• What it is: A mapping of African AI governance — what strategies say versus what states actually fund, build, and enforce.
• Why it matters: It stress-tests "strategy as a PDF" and pushes you to confront institutional capacity, budgets, and execution — where most strategies fail.
• Contribution: Supplies the sprint's realism: the implementation gap becomes a first-class topic from Week 1, not an afterthought.

Go deeper (optional)

• UNESCO: Recommendation on the Ethics of AI (2021)
• Why it matters: Adopted by 190+ member states, it is the most globally representative normative instrument on AI — broader in membership than the OECD baseline. Its companion Readiness Assessment Methodology (RAM) gives governments a concrete self-assessment tool you will meet again in Week 5.
• Use: Cite it when your context needs legitimacy grounded in a genuinely universal, Global-South-inclusive instrument rather than an OECD-club one.
• Oxford Insights: Government AI Readiness Index
• Why it matters: A country-comparative benchmark of government AI readiness across government, technology, and data-and-infrastructure dimensions — with strong Global South coverage.
• Use: Find your country's (or a comparable country's) score to anchor "where we stand" in a defensible external number.
• IMF: AI Preparedness Index (2024)
• Why it matters: Frames readiness through a macroeconomic lens — digital infrastructure, human capital, innovation, regulation, and labor-market impact — across income groups.
• Use: Pull one or two indicators to make the "cost of waiting" concrete in economic terms a finance ministry respects.

Key ideas to internalize

• A vision is a falsifiable claim about the future, not an aspiration. If it could belong to any country, rewrite it.
• Urgency lives in the cost of a year of delay — something specific you lose and cannot recover.
• Rule-maker versus rule-taker is the sharpest Global South framing of "why now."

Common failure modes

• Reaching for the aspiration before doing the evidence work.
• Borrowing another country's urgency ("the AI race") instead of naming your own exposure.
• Confusing a normative principle (OECD, UNESCO) with a strategy — principles set the guardrails; they do not decide the choices.

Exercise — The Postcard

Write a "postcard from 2030" in the voice of your chosen country, describing:

• One visible change AI has made to everyday citizen experience.
• One institutional change in government that made it possible.
• One regional or international shift where the country now has more voice.

Under the postcard, extract:

• Three concrete proof points that must be true by 2028 for the 2030 story to be credible.
• Three risks of doing nothing, each linked to one of this week's readings (AU, AI Index, Carnegie, or an optional source).

Live session (60 min)

Cohort share-outs → pattern synthesis → a "why now" drafting clinic. You will revisit this postcard in later weeks to watch your understanding deepen.

Carry-forward to your memo

Your postcard and proof points become the Bottom line and Context sections of the final memo (Pillar 1). Keep them; you will sharpen, not replace, them.

Week 2 — The architecture: learning the 9-pillar map

Goal: Learn how strong strategies are structured, and how context reshapes which pillars carry weight.

The idea

Once you have a vision, the next question is structure: what are the load-bearing parts of a strategy, and how do they fit together? The nine pillars are that structure. But the deepest lesson this week is that the same nine pillars look completely different depending on context — and that difference is information, not error.

A wealthy city-state with reliable power and deep capital will lean hard on compute and priority-sector systems. A large, mobile-first economy with fiscal constraints will lean on talent, data, and pragmatic sequencing. When a strategy is silent on a pillar, that silence usually reveals a context (a constraint, a political reality, a dependency) rather than an oversight. Reading strategies comparatively teaches you to distinguish a genuine gap from a deliberate choice — and to make your own choices deliberately.

The skill you are building is comparative diagnosis: run the nine-pillar map across several real strategies, mark each pillar as explicit / implicit / absent and high / medium / low emphasis, and then ask why. That "why" is where strategy judgment lives.

Anchor readings

• Singapore: National AI Strategy 2.0 (Executive Summary + systems approach)
• What it is: Singapore's updated strategy, structured around national "systems" and enabling foundations (talent, compute, data, governance).
• Why it matters: Widely cited as one of the most coherent and actionable strategies anywhere, with concrete actions, responsible agencies, and timelines.
• Contribution: A live example of Pillars 1–7 in action, each clearly distinguished — the gold standard for clarity and sequencing.
• Kenya: Artificial Intelligence Strategy 2025–2030
• What it is: Kenya's national strategy across enabling environment, skills, sectoral application, governance, and regional positioning.
• Why it matters: A flagship African case for balancing ambition against constraints — compute access, talent retention, mobile-first realities.
• Contribution: Your Africa-led design anchor. Use it to surface what changes when infrastructure and fiscal space are binding constraints.
• African Union: Continental AI Strategy (reference for regional alignment)
• ASEAN: Guide on AI Governance and Ethics (regional governance baseline)
• What it is: A regional guide (principles + practical guidance) that harmonizes approaches while allowing national variation.
• Why it matters: The best comparative example of "regional coordination without full legal harmonization" — directly relevant to AU ambitions.
• Contribution: A concrete model for how a region reduces fragmentation, shares capacity, and coordinates standards.
• How to use: Extract mechanisms — mutual recognition, shared testing norms, capacity support — that regional blocs can adapt.

Go deeper (optional)

• India: National Strategy for AI (#AIforAll) & the IndiaAI Mission
• Why it matters: A major Global South strategy that frames AI explicitly around social inclusion and scale ("AI for all"), then follows through with a funded national mission (compute, datasets, skilling, safety). A powerful third comparison point between Singapore's resource-rich model and Kenya's constraint-bound one.
• Use: Study how a large, diverse, budget-conscious democracy sequences compute and data as public goods.
• United Arab Emirates: National Strategy for AI 2031
• Why it matters: An ambitious, resource-rich small-state model with a dedicated ministry — a useful contrast that isolates what money and political will buy, and what they still cannot.
• Use: Compare its institutional machinery (a ministry, not a committee) against thinner arrangements elsewhere.

Key ideas to internalize

• Silence on a pillar is a signal about context, not proof of incompetence.
• Emphasis follows constraints: rich-and-reliable leans on compute; constrained-and-mobile leans on talent and data.
• Every strong strategy makes fewer promises than a weak one — it chooses.

Common failure modes

• Grading strategies against a single "ideal" instead of against their context.
• Mistaking length and polish for coherence.
• Copying the emphasis of a strategy written for a country unlike yours.

Exercise — Pillar mapping (1 page)

Map Singapore + Kenya + AU (add India or UAE if you like) against the nine pillars:

Then write three sentences: which omissions look like genuine gaps, which look like deliberate context-driven choices, and what that implies for your context.

Live session (60 min)

Facilitated live mapping → small-group mapping → report-outs. Expect disagreement about what counts as a "gap" — that disagreement is the lesson.

Carry-forward to your memo

Your pillar map becomes the backbone of the memo's Pillars and choices section (Pillars 1–8). Where your context's emphasis differs from the models, note why — that "why" is your strategy's argument.

Week 2.5 (Optional Deep Dive) — Constraints, leapfrogging, and realistic budgets

A single optional module — not a recurring half-week.

Goal: Practice strategy under real constraints: power reliability, limited compute, talent scarcity, donor dependence, and short political time horizons.

The idea

Ambition is cheap; feasibility is where strategies live or die. This module forces the honest question most strategies dodge: given our actual constraints, what can we credibly do in the next 18–24 months? Constraints are not only limits — they are also the source of leapfrogging. Mobile money, digital ID, and instant-payment rails let several countries skip infrastructure stages the West built over decades. The countries that leapfrogged did not have more resources; they sequenced around their binding constraint and built shared digital public infrastructure (DPI) that later became the foundation for AI.

The move to practice is identifying your binding constraint — the one that, if unaddressed, makes everything else impossible (often power or talent, sometimes fiscal space or political time). Strategy under realism means sequencing from that constraint outward, not pretending it away.

Anchor readings

• World Bank: Digital Progress and Trends Report 2025 (AI-foundations focus)
• What it is: A synthesis of digital-development progress, constraints, and policy levers relevant to DPI, connectivity, and public-sector modernization.
• Why it matters: A practical "state capacity + infrastructure" reference — exactly where AI strategies collapse if ignored.
• Contribution: Strengthens the foundation layer: DPI, service delivery, and institutional feasibility before the AI glamor.
• ASEAN: Expanded Guide (Generative AI supplement)
• Why it matters: Shows a region updating its guidance for generative AI without rewriting everything — a model for adaptive, incremental governance under constraint.
• CSIS — From Divide to Delivery: How AI Can Serve the Global South
• What it is: A Global South–focused analysis of AI opportunity, delivery constraints, and pathways to responsible deployment.
• Why it matters: It treats "AI for development" as implementation work — institutions, delivery, procurement, capacity — not demo theater.
• Contribution: A delivery-oriented counterweight to strategy documents that over-index on aspiration.

Go deeper (optional)

• Masakhane — grassroots African-language NLP
• Why it matters: A concrete answer to the data-and-talent constraint: a distributed, low-resource-first research community building language technology for African languages that big models neglect. Leapfrogging as a community and data strategy, not just an infrastructure one.
• Use: Cite it when your context's binding constraint is data scarcity or language coverage, not compute.
• India's DPI stack (Aadhaar, UPI, DigiLocker) as a leapfrogging reference — see the World Bank / G20 work on Digital Public Infrastructure
• Why it matters: The canonical example of sequencing shared rails (identity, payments, data exchange) first, which then made AI-enabled services cheap to build on top. The lesson is order of operations, not budget size.
• Use: Ask what your equivalent shared rails are, and whether they exist before you promise sector pilots.

Key ideas to internalize

• Find the binding constraint first; sequence everything else from it.
• Leapfrogging is a sequencing decision (skip a stage) enabled by shared rails, not a spending decision.
• Donor dependence and political time horizons are real constraints — budget for them explicitly.

Live session (90 min, optional)

An interactive comparison of the same strategy attempted under different constraint profiles — what survives contact with reality, and what has to be cut or resequenced.

Carry-forward to your memo

Your binding-constraint list disciplines the memo's Budget logic and Implementation plan — what gets funded first, and why the sequence is realistic.

Week 3 — The engines: compute, data, and institutions

Goal: Separate three commonly muddled topics — compute, data, and institutions — and design each realistically.

The idea

Three things get blurred in weak strategies, and untangling them is most of the work of Week 3.

Compute is physical and geopolitical: chips, data centers, connectivity, and — the pillar everyone forgets — energy. The strategic question is not "do we have GPUs" but "on whose terms do we access compute, and what happens to that access under stress?" Framed as strategic infrastructure rather than IT procurement, compute earns a place in cabinet-level budgeting.

Data is institutional and legal: quality, interoperability, privacy, and trusted-sharing arrangements. A country can have compute and still be unable to build anything useful because its public data is fragmented, unusable, or legally frozen. Compute and data are separate levers that require separate policy tools — conflating them is the single most common technical error in national strategies.

Institutions are the machinery that makes any of it happen. This is where most strategies quietly fail: they name a "coordinating council" with a mandate to coordinate and no power to make a ministry move. The design goal is the minimum viable institution — the smallest real thing with an owner, a budget, staff, and the authority to say no — plus a credible first-100-days.

Anchor readings

• Tony Blair Institute: compute/infrastructure as strategic asset (selected sections)
• What it is: An argument that compute and digital infrastructure function like strategic national infrastructure, shaping competitiveness and state capacity.
• Why it matters: It reframes compute from "IT spend" into "industrial + security infrastructure," which changes how it is budgeted and governed.
• Contribution: Sharpens Pillar 2 with an investment rationale policymakers recognize.
• How to use: Borrow its framing to justify why compute decisions belong in the strategy memo at all.
• World Bank: AI foundations and digital infrastructure (2025 report)
• Why it matters: A practical "state capacity + infrastructure" reference — where AI strategies collapse if ignored.
• How to use: Pull the constraints checklist (connectivity, payments/ID rails, government-data maturity) as prerequisites before any sector pilot.
• UK Government: Introducing the AI Safety Institute (institutional model)
• What it is: A description of the AISI's role, mandate, and rationale as a national AI-safety capability.
• Why it matters: It shows what a real institution looks like — scope, interfaces, credibility — not just "we will ensure safety."
• Contribution: Upgrades Pillars 6 and 7 from principles into implementable machinery.
• How to use: Treat it as a pattern library — testing, evaluation, incident response, cross-government coordination.
• African Union: Continental AI Strategy (data governance and coordination)

Go deeper (optional)

• NIST: AI Risk Management Framework (2023)
• Why it matters: The most widely adopted operational governance tool in the world — a voluntary, function-based framework (Govern, Map, Measure, Manage) that turns "we will govern AI" into concrete, auditable practice. It is the bridge from Pillar 6 principles to Pillar 7 machinery.
• Use: Use its four functions as the skeleton for your governance-and-oversight design instead of inventing one.
• The International Network of AI Safety Institutes — the emerging web of national safety institutes coordinating evaluations after the Bletchley and Seoul summits (see the UK AISI reading above as the entry point).
• Why it matters: Safety capacity is becoming a shared, joinable asset — a small country can plug into evaluation capability rather than build it alone.
• Use: Consider membership or association as a low-cost route to credible safety machinery.

Key ideas to internalize

• Compute and data are different levers needing different tools — never merge them in a strategy.
• Energy is part of the compute pillar. If the grid cannot support it, the compute plan is fiction.
• Design the minimum viable institution: an owner, a budget, staff, and the authority to say no.

Common failure modes

• A "national data strategy" that is really a privacy law with no plan to make data usable.
• A coordination body with responsibility but no authority.
• Safety framed as values, not as testing, evaluation, and incident response you can staff.

Exercise — Institutional wiring diagram (2 pages max)

Design a structure that could actually run:

• Strategy ownership — who leads, where it sits, and its reporting line.
• Delivery mechanism — how sector ministries execute, not merely "coordinate."
• A first-100-days plan — three concrete, dated deliverables.

Live session (60 min)

Institutional designs compared; failure modes named; the "minimum viable institution" principle stress-tested against each design.

Carry-forward to your memo

This wiring diagram becomes the memo's Institutions section (Pillar 7) and feeds the Ask and timeline — you now know what mandate and budget you are actually requesting.

Week 4 — Application: sectors, society, and international stance

Goal: Make the hard choices — priority sectors, labor and inclusion impacts, and international positioning.

The idea

Strategy becomes real when it chooses. Week 4 forces three choices that most strategies blur.

Sectors: "AI in health, education, and agriculture" is not a choice — it is a refusal to choose. The discipline is to name two or three sectors and, for each, one specific Year-1 use case with an owner and a definition of success. Everywhere means nowhere; the strategy that picks a beachhead is the one that ships.

Society and labor: AI reshapes tasks and jobs unevenly — some workers gain, some lose, and the distribution is not random. A strategy that treats labor as an afterthought loses public legitimacy the moment displacement starts. The move is to attach, to each priority sector, one concrete labor or inclusion risk and one credible mitigation (reskilling, transition support, complementary deployment that augments rather than replaces).

International stance: No country decides alone. Standards, export controls, model access, and regional frameworks constrain and enable every national choice. A deliberate posture names four things: regional alignment, vendor diversification, sovereignty red lines, and standards participation. The alternative — passive alignment by default — means importing someone else's constraints without noticing.

Anchor readings

• Singapore NAIS 2.0: sector-systems sections (how real strategies pick use cases)
• Why it matters: Demonstrates what "implementation-ready" looks like — key systems, responsible actors, coherent sequencing.
• Use: Model your own use-case selection on its specificity, not its scale.
• MIT ICEBERG: work and skills impacts (overview hub)
• What it is: A research platform tracking how AI changes tasks, jobs, and skill demand.
• Why it matters: It helps you stop guessing about labor impact and start structuring it — who gains, who loses, where, and when.
• Contribution: Makes the future of work a measurable pillar with design implications for training and social protection.
• Use: Translate its insights into one concrete labor mitigation per priority sector.
• AU Continental AI Strategy: sovereignty and development emphasis
• AI geopolitics and international coordination — two short Carnegie pieces:
• The AI Governance Arms Race — From Summit Pageantry to Progress
• In Which Areas of Technical AI Safety Could Geopolitical Rivals Cooperate?

Go deeper (optional)

• EU AI Act
• Why it matters: The world's first comprehensive, horizontal AI law, built on risk tiers. Whatever your context, its "Brussels effect" will shape the standards your exporters and vendors must meet — so your international posture has to have a position on it.
• Use: Decide whether you align with, adapt from, or deliberately diverge from the risk-tier model — and say so explicitly.
• Bletchley Declaration (2023) and the Seoul follow-up (2024)
• Why it matters: The frontier-safety summit track is where the international norm-setting is actually happening. Presence at that table is itself a strategic choice.
• Use: Treat participation (or association) as a line item in your international-engagement pillar.
• ILO: Generative AI and Jobs (2023)
• Why it matters: Models exposure to generative AI by task and by income group, with an explicit Global South cut — augmentation is far more likely than wholesale automation in many economies.
• Use: Ground your labor-mitigation choices in who is actually exposed in your context, not in headlines.

Key ideas to internalize

• Two or three sectors, each with one named Year-1 use case. Choosing is the strategy.
• Attach a labor risk and a mitigation to every sector, or lose public legitimacy later.
• A real international posture names its red lines — the things you will not concede.

Common failure modes

• Listing sectors instead of choosing use cases.
• Treating labor impact as a communications problem rather than a design input.
• Assuming neutrality is a posture. Default alignment is a choice you did not make on purpose.

Exercise — Priority and posture (1–2 pages)

For your chosen context:

• Select 2–3 priority sectors and define one specific Year-1 use case per sector (not "AI in health").
• Identify one labor/inclusion risk per sector and one mitigation (skills, transition support, job creation, or complementary deployment).
• Write a short international-stance paragraph: regional alignment + vendor diversification + sovereignty red lines + standards participation.

Live session (60 min)

Sector choices justified out loud; international stances stress-tested by peers playing skeptical ministers.

Carry-forward to your memo

These become the memo's Priority sectors (Pillar 5) and International engagement (Pillar 8) sections — the most concrete, and often most persuasive, parts of the final document.

Week 5 — Synthesis: metrics, iteration, and the strategy memo

Goal: Turn everything into a decision-ready memo and a short spoken briefing — an adaptive strategy, not a static document.

The idea

Two disciplines converge in the final week.

Metrics decide whether a strategy is adaptive or dead on arrival. The trap is vanity metrics — strategies published, workshops held, MOUs signed — which measure activity, not outcomes. Good metrics are few (three to five), they have named owners and a review cadence, and each one has a pivot point: a threshold that, if crossed the wrong way, forces a change of plan. A metric that cannot change your behavior is decoration.

Format decides whether anyone acts on your work. Senior leaders decide from one- to two-page briefs, not fifty-page reports. The memo discipline — bottom line first, choices and trade-offs in the middle, a clear ask at the end — is not bureaucratic ritual; it is respect for the reader's attention and the fastest path from analysis to decision. The three-minute pitch is the same discipline compressed further: if you cannot say it in three minutes, you have not finished deciding.

Anchor readings

• OECD.AI: monitoring national AI policies and capability indicators
• What it is: Dashboards and reports tracking how countries put AI principles into practice and how capabilities evolve.
• Why it matters: Shows how indicators can track progress and inform course-correction.
• Contribution: Direct support for Pillar 9 (monitoring, metrics, review).
• Harvard Kennedy School: How to Write a Policy Memo (format discipline)
• Why it matters: Leaders decide from one- to two-page briefs. This gives you a clean, reusable format.
• Contribution: The template for your final deliverable and for future work.
• World Bank 2025 report (metrics and foundations framing)

Go deeper (optional)

• UNESCO: Readiness Assessment Methodology (RAM)
• Why it matters: A structured, government-facing tool for assessing where a country stands against the UNESCO Recommendation — a ready-made baseline-and-review instrument you can adopt rather than build.
• Use: Borrow its dimensions as your review checklist so your metrics connect to an international baseline.
• Stanford HAI: AI Index — metrics chapters
• Use: Return to it for indicator definitions and benchmarks so your three-to-five metrics are comparable to how others measure the same thing.

Week-Five deliverables

1. Two-page National AI Strategy memo (decision-ready):

• Bottom line — the clear ask.
• Context — the "why now," grounded in reality.
• Strategy summary across the nine pillars — tight, not exhaustive.
• Implementation plan — Year 1–2 in detail; Year 3–5 in direction.
• Budget logic — what is funded first, and why.
• Risks + mitigations + pivot points.
• Metrics — three to five that drive behavior, not vanity.

2. Three-minute oral pitch:

• 30 seconds — why now.
• 60 seconds — top strategic choices.
• 60 seconds — implementation and what you need approved.
• 30 seconds — a memorable, un-inflated closing line.

Live session (90 min)

Memo workshop → structured peer review → pitch practice, with the room playing the decision-maker.

Carry-forward

This is the carry-forward. Every fragment you emitted in Weeks 1–4 assembles here. If the earlier weeks were done well, Week 5 is composition, not creation.

Final deliverables

You will write a two-page memo to a head of government, CEO, or comparable decision-maker (real or fictional), structured as follows:

1. Bottom line (Vision & narrative). One or two sentences: what your country wants from AI, and why now.
2. Pillars and choices (1–8). Short sections covering: compute and infrastructure (Pillar 2); data and DPI (Pillar 3); talent and R&D (Pillar 4); two or three priority sectors and public-sector transformation (Pillar 5); governance and rights guardrails (Pillar 6); the institutions that will run and coordinate the strategy (Pillar 7); and international-engagement stance (Pillar 8).
3. Metrics and review (Pillar 9). Three to five indicators for the next 3–5 years, who reports them, and how often the strategy is reviewed.
4. Ask and timeline. The decisions, budgets, or mandates you are requesting now, and what is realistically achievable in 12–24 months.

You will also deliver a three-minute oral pitch, practicing how to explain complex trade-offs in plain language.

What "good" looks like (memo rubric)

Use this to self-assess before peer review. A strong memo:

• Chooses. It could not have been written about any other country. Names, sectors, and constraints are specific.
• Sequences. It says what happens first and why, not just what should exist eventually.
• Owns. Every major move has a named owner with the authority to deliver it.
• Budgets honestly. It funds the binding constraint first and admits what it is not doing yet.
• Measures behavior. Its metrics have owners, cadence, and pivot points — not vanity counts.
• Takes a posture. It states at least one international red line it will not concede.
• Fits on two pages. Discipline of length is discipline of thought.

What makes this sprint different

• Context-first design. Infrastructure gaps, power constraints, and limited compute access are first-order variables, not footnotes.
• Execution emphasis. Institutions and delivery machinery are treated as the core of strategy, not an appendix.
• Regional realism. National strategy is aligned with regional frameworks (AU, ASEAN) without collapsing national agency.
• Modern governance coverage. Safety capacity, audits, and oversight are buildable components, scaled to budget reality.
• Decision-ready outputs. The program forces clarity, sequencing, and trade-offs — artifacts leaders can act on.

Learning outcomes

By the end of the sprint, you should be able to:

• Use the nine-pillar framework to analyze and compare national or organizational AI strategies.
• Distinguish clearly between compute and data as separate strategic levers, and design policy tools for each.
• Identify and justify priority sectors and matching public-sector transformation plans.
• Propose realistic institutional arrangements for implementation, including coordination and oversight bodies.
• Articulate a clear international posture, aligned with existing global frameworks and geopolitical realities.
• Embed labor-market, inclusion, and societal impacts into strategy rather than treating them as afterthoughts.
• Define meaningful metrics and review cycles that keep strategy adaptive rather than static.
• Communicate all of the above in a short, persuasive memo and spoken briefing suited to senior decision-makers.

The reading library (organized by pillar)

Every source used in the sprint, mapped to the pillars it most informs — a reference you can return to long after the cohort ends.

Program summary

The CentPol National AI Strategy Sprint (2026) is a five-week, hybrid program that helps policymakers and strategists design implementable national AI strategies for Africa and the Global South. Participants learn a reusable nine-pillar framework — vision, compute, data governance, talent, priority sectors, governance and safety, institutions, international engagement, and metrics — and apply it week by week to progressively harder layers of a real strategy. Each week combines curated public readings, live discussion, and a structured exercise that emits a fragment of the final deliverable. By the end, participants produce a two-page strategy memo and a three-minute pitch built for real decision-makers.

About CentPol

CentPol (Center for Emerging and Next Tech Policy, Strategy & Foresight) helps governments and institutions make sense of fast-moving technologies and design strategies that are ambitious, realistic, and responsible. Its work draws on comparative strategy analysis, regional governance frameworks, and implementation-focused policy design.