What looks like responsiveness and adaptation as Ghanaian public institutions increasingly purchase automated software and artificial intelligence solutions to digitalize operations, is actually stalled inside committees who lack the technical expertise to evaluate algorithmic risk or data constraints. These decisions are highly evident in the procurement of automated systems in sectors like health and education, critical in every modern economy.

The consequence of a committee made up of a head of procurement unit, financial controller, an end user and an internal IT administrator—a technical literacy gap transcends bad purchase or misallocation of funds. Instead, dysfunctional systems and exploitative vendor lock-ins are deployed into critical systems that threaten the stability of operations. Because even advanced countries like the UK lack clear audit standards and a universally accepted framework, integrating unaudited ones into essential sectors like healthcare diagnostics or school grading systems introduces huge loopholes into ethical, clinical and legal frameworks.

More concerning still, there are no common international guidelines too. However, public sectors deploy systems developed outside their borders exposing them to predatory technology practices. With no implication ever, vendors are mostly guided by how fast they implement, at the expense of the safety, sustainability and the digital sovereignty of the buyer. An attempt to make these systems reliable likely requires transparency and proactive governance of personal digital data such systems use in decision-making. Under discussions are systems that sometimes unintentionally might cause societal biases and violate fundamental human rights.

This introduces why committees operating and signing public procurement decisions must possess the technical expertise since they decide a nation’s digital sovereignty. For a procurement law like Ghana’s Public Procurement Act (663, amended by Act 914), assessing the algorithmic tools that are fast evolving is a hurdle since its provisions are originally designed for standardized physical goods. The provision mainly guides cost and history guidelines evaluation matrices. The standard document for IT products is also fundamentally inadequate. It is designed to cover traditional installation schedules and maintenance routines. Therefore, adopting such a document in evaluating the algorithmic logic of systems and even properly handling data creates critical bias and data risk that cannot be ignored.

The primary beneficiaries of this procedural rigidity are technology vendors who present their software capabilities as more than they are, manipulating the bidding process. This practice of ‘AI washing’ is further enabled by committees lacking computational literacy. As such, basic rule-based systems, often presented as sophisticated, are purchased at outrageous cost. The vendor’s playbook is to submit a polished Request for proposal and have a pitch deck that promises the ideal. More often than not, the aftermath is a mismatch of what is specified and reality since they barely help solve a problem.

With a procurement framework unintentionally structured to control competition in favour of foreign managed systems, localised GovTech startups get disqualified early in the process. The latter who might have built systems offering solutions tailored to local problems are discouraged from advancing in the bidding process. However, their foreign counterparts seemingly proceed since they have the resources to navigate complex bureaucratic RFP requirements.

Interestingly, there are no clear provisions on the governance of digital infrastructure and long term continuity protocols. This somewhat implies how vendor contracts are likely to omit clauses on the state’s ownership of data or even an access to the source code. In this case, a country’s digital sovereignty is left hanging.

To address this, dedicated algorithmic audit frameworks are needed to ensure the procurement of automated systems are fair, robust and legal. Relying entirely on the already existing ones for purchase decisions might not be supported by any proper inspection of algorithms. Beyond that are also consequences related to after-purchase. In the case where systems break down or expire, transfer to the state becomes almost impossible. These procurements are mostly done without a dedicated policy or any legal mandate to ensure the smooth transfer of systems. What this means for a government is a digital sovereignty at risk, depending on private entities for newer solutions.

The inspection and verification of these systems shouldn’t be left trapped in the four corners of a board room. Rather, it would require the engagement of its users. Affected communities likely to benefit from the system should participate in designing and overseeing systems that directly impact them. Coupling this requires external ethic committees and third-party algorithmic inspections will ensure no unverified assumptions are programmed into critical public sector AI. This then exhibits true algorithmic accountability that considers participatory data stewardship.

Award of contract or purchase of a software shouldn’t be based entirely on its upfront cost. Costs could come in the form of system limitation and any logic flaws or improper training data acquisitions. So, before a decision is made, the disclosure of information pertaining to the source of data or any limitations of the automated platform must be mandatory even before it is used. In a bid to ensure that technical vetting standards are enforced, the “ethics by design” approach can be adopted in the initial phase of the bidding process. The approach would encourage the ethical design and deployment of AI systems. This would also mean that unverified automated logic would not be integrated into public systems easily.

Rapidly, the consequences are being faced in areas as critical as the education sector in Ghana. Every year, thousands of anxious parents assemble at national resolution centres for their wards to be placed in Senior High Schools while the system meant to automatically make placements is adjusted. The computerized School Selection and Placement System (CSSPS) operates on a rigid algorithm that doesn’t consider the differences in local school facility funding, thereby unable to interpret nuanced human flaws or correct wrong data. A typical placement cycle sees placement of more than required number of students to a category “A” school or glitched allocations such as male students placed in an all-girls school. To solve this, decentralized, real time feedback mechanisms are urgently required.

A more recent one experienced is the rollout of the Lightwave Health Information management systems (LHIMS). For a system used in a public agency as critical as the healthcare sector, its rollout revealed the state of the nation’s digital sovereignty. Major hospitals who had almost the entire sensitive health data for millions of citizens returned to the manual paper for record keeping. This was because the private vendor still had exclusive control of a system so delicate for a nation.

A keen view of the consequences of reliance on monolithic vendor platforms, is a pregnant woman who relocates and transfers her antenatal care to a regional hospital in another city. With the rollout, her medical history becomes untraceable, leaving health professionals to rely on guess work.

For these reasons, public agencies seeking to modernize operations should not leave purchasing decisions solely in the hands of procurement boards and committees operating with a technical literacy gap. They do not just sign-off unverified algorithms but are signing away the nation’s digital sovereignty.

Key Takeaways

• Public procurement committees should include independent computational linguists and data ethics experts to evaluate proposals before awarding contracts. Having these technical experts on board helps fish out any acts of ‘AI washing’ and unverified transformative claims.
• Technology contracts must explicitly guarantee national data sovereignty by mandating local server hosting and ensuring state access to the underlying source code. Clauses must be provided stating clearly digital exit strategies and interoperability standards legally required to prevent costly, exploitative vendor lock-in.
• To ensure that localized, agile GovTech developers are competitive, government procurement strategies should heavily mandate open-source capabilities, devoid of large, rigid system integrators. This might likely welcome tailored solutions.