The Ghana Revenue Authority (GRA) has introduced a new artificial intelligence system called Publican AI. This system helps with valuing and classifying imported goods. Since its launch in April 2026, it has caused some concern among traders and freight forwarders. They worried that the AI would make all the decisions. This created a tense atmosphere in Ghana's trading community. Declarations that used to go through normal channels now faced 'algorithmic resistance.'
The main issue was not the use of AI itself. Many countries use AI to manage trade risks. The problem was the uncertainty. Traders did not understand how Publican AI worked. They did not know what made it change values so much. Some early messages from within the GRA suggested the AI's decisions were final. This made it seem like customs officers could no longer use their own judgment. They thought traditional checks on documents were replaced by computer-generated results. Later, the GRA provided more information. This helped to change that perception. They explained that the system is meant to help officers. It supports their decisions and flags goods needing closer checks.
This difference in understanding is very important. The debate is not just about AI versus traders. It is about how the system's logic is actually used. The A-B-C corridor model helps explain this. Publican AI works using a 'confidence corridor.' It does not always look for one single exact value. Instead, it checks if a declared value falls within a range of what is commercially believable. This range is based on past trade patterns. It also considers how well documents match and how traders normally behave.
The A-B-C model uses three main points. Point A is the lowest possible value. Point C is the highest possible value. Point B is the most likely value. If a trader's declared value is between A and C, the system sees it as commercially believable. This means the declaration does not break the system's rules. Within this corridor, the customs officer can still make decisions. The officer might accept the value. They might ask for small changes. Or they could adjust the value a bit. The main thing is that the declaration can still be explained. The system basically says: 'This declared value is in a range that makes sense for business.'
This is a key point because customs AI is not like a simple calculator. It is more like an engine that builds confidence. It looks at many things. These include past trades, how well all the papers fit together. It checks if the declared category for goods is right. It also considers payment methods and the seller's history. The system works using probabilities and evidence. It does not aim for one universal price. This is vital in international trade. Business deals are rarely exactly the same.
The tension after the system started was mainly due to how people saw it. Many traders thought the AI would set the value. They believed officers could not change the AI's number. They thought their own assessments did not matter anymore. This caused fear. Customs valuation relies on checking transaction values and documents. It does not rely on AI setting prices. Early results seemed to confirm these fears. Traders believed their values were fair. However, these values were suddenly changed upwards. This happened in ways that seemed sudden and not well-explained. Because the market did not understand the confidence-corridor idea, many saw the changes as random. The GRA's later clarification was very important. It reminded everyone that Publican AI is there to help them decide. It is not meant to remove customs officers' ability to judge.