Ghana’s AI strategy is a laudable list of operating imperatives, not a strategy.

Ghana’s newly launched National Artificial Intelligence strategy deserves credit for treating AI as a national development question rather than a narrow technology policy issue. It is one of the more ambitious public AI documents to emerge from Africa, linking artificial intelligence to agriculture, healthcare, jobs, public-sector reform, language inclusion, and long-term economic transformation. It is also ethically alert, explicitly aligning itself with the UNESCO recommendation on the ethics of Artificial Intelligence and proposing a Responsible AI Authority to oversee implementation.

That is the good news. The harder truth is that, judged through Roger Martin’s strategy cascade lens, Ghana’s AI strategy often reads less like a true strategy and more like what Martin calls a “laudable list” — a long catalogue of worthy initiatives that signals seriousness, but does not yet make enough hard choices about where Ghana will play and how it will actually win.

That distinction matters. Countries, like companies, do not win because they have the longest to-do list. They win because they make a small number of integrated choices that reinforce one another and create an advantage others cannot easily match. Ghana’s AI strategy shows real ambition. The question is whether it shows enough strategy.

Big vision, loose definition

The strategy’s mission is to “harness AI for inclusive growth across all sectors” and to position Ghana as “a trailblazer for AI leadership in Africa and beyond.” Its vision is “Ghana 2035: The AI-Powered Society,” with the country becoming “the leading African AI hub.”

This is bold and politically useful language. It signals confidence, national seriousness, and a welcome refusal to treat Ghana as merely a consumer of imported tools. But Roger Martin warns that weak strategies often begin with aspirations that are either too vague or too grand to guide real choice. “Leadership,” “trailblazer,” and “hub” sound impressive, but they are not strategic unless they specify what winning actually means and relative to whom.

That is where the Ghana document becomes slippery. Does winning mean becoming Africa’s top destination for AI investment? The best home for public-sector AI? The leading centre for African-language models? The most effective adopter of AI in agriculture and health? The strategy points to all these possibilities but does not choose among them. In Martin’s terms, the aspiration is energizing but not yet discriminating enough to discipline the choices that follow.

A powerful document with too many arenas

The document is strongest when it recognizes that AI is not just about frontier models, but about development outcomes. It highlights practical use cases in agriculture, healthcare, environmental monitoring, transport, public administration, and culture. It proposes interventions across education, jobs, infrastructure, data governance, research, ecosystem building, private-sector adoption, and state capacity.

Yet this breadth is also the Strategy’s central weakness. Roger Martin’s framework insists that strategy requires a clear “where to play” choice — a defined set of arenas in which advantage will be pursued. Ghana’s AI Strategy, by contrast, tries to play almost everywhere: all sectors, all regions, all major layers of the AI stack, from foundational research and compute to adoption, regulation, start-up finance, and public procurement.

The problem is not that any of these priorities are wrong. Most are sensible. The problem is that strategy is not the art of compiling all sensible things into one deck. It is the art of deciding which few arenas matter most, because resources, institutional bandwidth, political attention, and implementation capacity are finite.

A national strategy that says “yes” to nearly every plausible AI opportunity risk becoming what Martin describes as “playing better” rather than “playing to win.” More hubs, more fellowships, more pilots, more datasets, more institutions — these may all improve the ecosystem somewhat. But improvement is not the same as competitive advantage.

The missing theory of advantage

Martin’s most useful test is simple: does the strategy contain a theory of advantage? In other words, does it explain why this actor — in this case Ghana — will be better than relevant competitors in a specific field of play?

The Ghana strategy contains many mechanisms. It proposes a National AI Fund beginning with 5 billion Ghana cedis, rising to 15 billion over the following period. It targets 200 billion Ghana cedis in foreign and local private investment and projects AI could contribute 500 billion Ghana cedis to GDP by 2035. It calls for a National Deep Science Institute, an NLP Centre of Excellence, national data repositories, large-scale compute expansion, AI fellowships, a Ghana Global AI Summit, AI use-case pilots, and even “GhanaChat,” a government-facing language model.

All of that signals energy. But energy is not a theory of advantage. Martin warns that the weakest “how to win” choices are usually lists of initiatives masquerading as strategy. The question is not whether the country should train people, improve infrastructure, collect data, or support innovators. Of course it should. The real question is: what combination of choices will make Ghana distinctively better than peers such as Rwanda, Kenya, South Africa, Egypt, or Morocco in specific AI domains?

That answer is never made explicit. The document gestures toward several plausible advantages — a growing innovation ecosystem, existing digital-policy foundations, Google’s AI Centre in Accra, local firms such as Farmerline and MinoHealth, and strong relevance in agriculture, public service delivery, and language technologies. But it does not turn those assets into a small number of tightly argued plays.

A more strategic version of the document might have said, for example: Ghana will aim to become Africa’s leader in three specific domains — climate-smart agriculture AI, African-language and public-service language technologies, and responsible AI implementation in government. It could then have explained exactly how Ghana’s data systems, research institutions, procurement rules, start-up finance, local language assets, and public-sector reform agenda would combine to create a reinforcing advantage in those three areas. That would be closer to strategy.

Pixels, portraits, and the danger of policy clutter

One of Martin’s most vivid metaphors is that weak strategies are made of “pixels,” while strong strategies form a “portrait.” A pixel is a sensible stand-alone activity. A portrait is a coherent picture showing how those activities fit together to produce a win.

Seen this way, Ghana’s AI strategy is full of impressive pixels. There is AI-ready training, youth reskilling, regional innovation hubs, cloud partnerships, open-data initiatives, an upgraded data-protection regime, AI research centres, sector pilots, public procurement reform, and the Responsible AI Authority. Almost every one of these is defensible. But the portrait remains blurry.

For instance, take agriculture. The document rightly identifies AI use cases such as increasing crop yields, soil management, farm management, and food-fraud detection. Ghana could build a genuinely distinctive agricultural AI play by linking farmer-facing platforms, localized agronomic data, weather intelligence, credit scoring, extension services, research institutions, and public procurement for food systems. But the strategy does not yet tell that story as an integrated system. The reader sees the nodes, but not the links.

The same problem appears in health. The strategy points to AI-enabled diagnostics and productivity gains in healthcare, which is sensible in a country where access, specialist shortages, and uneven service quality remain real constraints. Yet it stops short of defining a specific Ghanaian health-AI advantage — for example, becoming West Africa’s leader in AI tools for triage, radiology support, maternal risk prediction, or district-level public health surveillance through a blend of local data governance, provider partnerships, and regulatory credibility. Again, there are pixels. The portrait is not yet painted.

Operating imperatives are not strategy

Martin makes another important distinction: some choices are not strategy at all, but “operating imperatives.” If the opposite of a choice is obviously stupid, then it is not strategic. It is simply something any serious organization must do.

This insight sharpens the critique of the Ghana document. “Improve digital infrastructure.” “Promote ethical AI.” “Strengthen data governance.” “Expand skills.” “Support innovation.” “Include women, youth, and marginalized communities.” All of these are necessary. None are controversial. But precisely for that reason, they are not where strategy lives.

A country does not gain advantage merely by embracing operating imperatives. Every serious government in the AI era should be trying to improve infrastructure, talent, governance, and inclusion. The strategic question is what distinctive choices Ghana will make on top of those basics — what it will prioritize, sequence, and perhaps deliberately not do.

That “what not to do” question is largely absent from the Strategy. There is no strong sense of exclusion, sequencing, or refusal. But strategy without exclusion is usually just planning.

The implementation risk is overextension

To its credit, the document does not ignore implementation. It proposes a Responsible AI Authority, ongoing monitoring, public-sector coordination, and stronger enforcement capacity through an upgraded data-protection regime. Those are serious institutional moves.

Still, the implementation challenge is not only bureaucratic. It is strategic. Martin argues that the final test of a strategy is whether the organization has the capabilities and management systems to sustain the advantage it claims. On this front, Ghana’s Strategy asks an extraordinary amount of the state, universities, regulators, the private sector, and development partners all at once.

There are compute targets, energy targets, jobs targets, investment targets, dataset targets, unicorn targets, institutional reform targets, and wide-ranging adoption targets across multiple sectors. Such ambition can inspire. It can also overload the system. If everything is a priority, implementation fragments, attention thins out, and the document risks becoming a symbol of intent rather than a tool of disciplined execution.

This is why Martin’s framework is valuable for public policy: it reminds policymakers that execution problems often begin as choice problems. A strategy becomes hard to implement not only because institutions are weak, but because the underlying design asked them to do too many loosely connected things at once.

What a sharper Ghana AI strategy would look like

A stronger version of this Strategy would not abandon its developmental ambition. It would narrow and sequence it. It would define a handful of fields where Ghana has a realistic shot at continental distinction and then align finance, regulation, talent, research, procurement, and diplomacy around those fields.

Three possibilities stand out from the document itself. First, AI for agriculture and climate resilience, where Ghana’s development needs are urgent and measurable, and where local data and advisory systems could create real productivity gains. Second, African-language and public-service AI, where the combination of language inclusion, government digitization, and national data assets could produce tools with regional relevance. Third, responsible AI governance and public-sector deployment, where Ghana could aim not only to regulate well, but to become Africa’s model for trustworthy, effective state use of AI.

These are not the only possible bets. But they are the kind of bets a strategy must make. They would force clearer trade-offs. They would make investment discipline easier. They would clarify what the Responsible AI Authority is actually trying to optimize. And they would give the rest of the ecosystem — universities, founders, investors, civil society, and development partners — a more legible national signal.

The real opportunity

Ghana’s AI Strategy should not be dismissed. It is substantial, thoughtful, and far ahead of the complacency that still characterizes digital policy in many places. It understands that AI can shape productivity, public services, and inclusion. It also understands that ethics, data governance, and institutional design are not side issues, but central ones.

But the document would become much more powerful if it absorbed Roger Martin’s most important lesson: strategy is not a list of admirable intentions. It is an integrated set of choices about where to play and how to win, backed by capabilities and systems that make those choices real.

Right now, Ghana has an AI agenda with significant promise. To turn that agenda into a winning national strategy, it must make fewer promises, sharper choices, and more explicit claims about the advantage it seeks to build. That is not a call to lower ambition. It is a call to give ambition the discipline it needs to succeed.

Author

Kwame Gyamfi, Analyst