- Ghana Must Turn AI Ambition into a Productivity Engine Inside the State — Seth Terkper
Ghana’s artificial intelligence ambitions face a test that has less to do with the sophistication of algorithms and more to do with whether the technology can make government measurably more productive, according to Seth Terkper, Economic Adviser to the President. He has called for AI to be integrated into the public-sector systems already used to administer taxation, expenditure, payments and other government services. The argument shifts the country’s AI debate from technology adoption towards measurable economic returns.
Mr Terkper’s proposition is significant because Ghana has already invested substantially in digitising parts of public administration, including tax collection, financial management, identification and public-service delivery. Rather than creating entirely new technological structures, AI could operate as an intelligence layer over existing systems, allowing government to extract more value from information it already collects. The challenge would therefore be to move from digital government towards systems capable of analysing patterns, anticipating risks and supporting better decisions.
A digitised tax platform, for example, can record transactions and declarations, while an AI-assisted system could potentially identify unusual relationships between declared income, imports, corporate activity and other available information. That could allow tax authorities to concentrate audits and enforcement resources on cases presenting higher risks rather than relying predominantly on broad manual review. Similar applications could extend to customs, procurement, social protection and public expenditure management.
The fiscal implications could be considerable for a government seeking to increase domestic revenue without relying continually on higher tax rates. If AI-assisted analysis improves the detection of under-declaration or non-compliance, government could raise revenue through better enforcement, while automation of repetitive administrative processes could reduce the time and cost involved in delivering public services. Analytical systems could also help identify unusual procurement or expenditure patterns earlier, potentially allowing officials to intervene before losses become entrenched.
But the potential benefits are not automatic, and the quality of the underlying systems will determine whether AI improves government performance or merely adds another layer of technology. Poor-quality data can generate inaccurate recommendations at scale, while fragmented databases can prevent systems from detecting useful relationships across institutions. Weak controls over sensitive government information could also create significant cybersecurity and privacy risks.
Government data itself is increasingly becoming a strategic economic asset. Tax records contain information about economic activity, customs systems capture trade flows, procurement platforms record contracts and suppliers, while financial-management and social-protection systems hold large quantities of expenditure and household information. Connecting those datasets responsibly could allow AI to flag suspicious transactions, identify duplicate beneficiaries, forecast demand for public services and provide policymakers with faster assessments of emerging trends.
The same integration, however, would increase the consequences of a major data breach or system failure. A more intelligent government therefore requires stronger cybersecurity, audit mechanisms, access controls, privacy protections and rules governing how information is shared across institutions. AI policy cannot be separated from institutional governance because the usefulness of advanced analytical systems depends on the integrity of the data and controls surrounding them.
There is also a risk that government could fall into what amounts to an AI procurement trap purchasing sophisticated systems before addressing the organisational weaknesses that limit their usefulness. Technology cannot by itself repair inconsistent data standards, poorly designed workflows or institutions that are unable to exchange information effectively. An advanced AI platform layered over an inefficient bureaucracy could therefore become an expensive technological addition without materially changing how the state operates.
A more productive approach would begin with clearly defined government problems and then determine whether AI can solve them. Tax compliance, customs risk assessment, procurement monitoring, fraud detection and public-service automation offer possible starting points because their impact can potentially be measured in financial or operational terms. The key question should therefore not be how much AI government has deployed, but what specific outcomes the technology has improved.
That would require government to establish performance indicators capable of measuring the return on public investment in AI. Additional revenue collected, procurement losses prevented, working hours saved, faster processing times and reductions in the cost of delivering public services could all provide more useful measures of success than the number of platforms purchased. Such metrics would also be important in a fiscally constrained environment where every cedi spent on technology competes with other public priorities.
Human oversight would remain essential, particularly where automated assessments influence decisions affecting citizens directly. AI-generated recommendations in areas such as taxation, welfare, procurement, law enforcement or access to public services should not become substitutes for accountable administrative decision-making. Citizens should also retain meaningful avenues to challenge decisions in which automated systems have played a significant role.
Mr Terkper’s intervention ultimately points to a more pragmatic way of measuring Ghana’s AI ambitions. The country does not necessarily need the continent’s largest AI ecosystem to derive significant economic value from the technology; it needs to become effective at applying it to the inefficiencies that already impose costs on the state and the wider economy. That means extracting greater productivity from existing digital infrastructure rather than repeatedly constructing parallel systems.
If implemented effectively, AI could help government collect more revenue, reduce waste, improve service delivery and make public institutions more responsive, but doing so will require interoperable databases, reliable information, strong cybersecurity and rigorous measurement of results. The central challenge is therefore no longer simply whether Ghana should adopt AI, but whether it can embed the technology deeply enough into the machinery of government to generate a measurable economic return. In that sense, Mr Terkper’s call for Ghana to “integrate AI into existing government systems” is ultimately a productivity agenda rather than merely a technology strategy.
