Ghana’s National AI Strategy sets a goal of curating 1 trillion tokens of local data by 2030, a target that depends on businesses keeping cleaner records than many currently do.

President John Dramani Mahama launched Ghana’s National AI Strategy on April 24 at the Labadi Beach Hotel in Accra. The plan spans 2025 to 2035 and rests on eight pillars, backed by a National AI Fund that starts at 5 billion cedis, about 450 million US dollars, through 2030 and scales to 15 billion cedis by 2035. Officials want AI to add 200 billion cedis to GDP by 2030 and 500 billion cedis by 2035.

One of the eight pillars covers data access and governance. It sets a target of curating 1 trillion tokens of Ghanaian datasets by 2030 and proposes grants to fund dataset creation and annotation in health, agriculture and education. A Responsible AI Authority is meant to coordinate all eight pillars within the strategy’s first year.

That data target depends on information that does not yet exist in usable form inside most Ghanaian businesses. Many small and midsized companies still keep customer records in notebooks or scattered spreadsheets, assign overlapping duties without written job descriptions, and rely on the owner’s memory rather than documented processes.

TechTarget enterprise software editor James Alan Miller made a similar point about large companies adopting AI tools. “AI didn’t break enterprise systems, it exposed them,” he wrote.

The same pattern plays out at a much smaller scale. Software that depends on consistent records will surface whatever memory and improvisation have quietly patched over, whether inside a multinational’s resource planning system or a market trader’s spreadsheet.

Recent industry research points to the same conclusion. A Confluent study reported by TechRadar found that most organisations do not have an AI investment problem so much as a data problem, with about 72 percent of IT leaders blaming poor data infrastructure, rather than a shortage of money or tools, for stalled AI projects.

For Ghanaian SMEs weighing whether to adopt AI tools for customer service or data analysis, that finding suggests the more useful first step may be auditing existing records rather than shopping for new software. AI systems built on inconsistent or duplicated data tend to produce unreliable results, regardless of how advanced the underlying model is.

Whether Ghana reaches its own 1 trillion token target by 2030 may depend less on the size of the National AI Fund than on how many individual businesses sort out their basic records long before any AI tool touches them.