The Kenya Revenue Authority will launch what could become one of the country’s most advanced uses of Artificial Intelligence — an AI-powered cargo scanning system designed to seal loopholes, boost efficiency and outsmart smugglers.
KRA is piloting advanced AI systems that promise to revolutionise cargo inspection, risk profiling and revenue assurance at all ports of entry.
A date for full operation has not been disclosed.
Unlike traditional image analysis systems, KRA's new AI is being trained to do much more than just look at X-ray images.
The system is being trained to perform consignment analysis — not just identifying what’s inside a container, but also determining whether a shipment’s declared value matches known trade data, assessing importer risk profiles and flagging inconsistencies with ruthless accuracy.
KRA’s Albert Atambo said the agency is exploring means to deploy AI to manage its scanners at borders. He is the chief manager of Customs and border control, Scanner management section.
He said recognising both the complexity and cost of implementing such a system, KRA has taken a cautious, evidence-based approach.
A proof-of-concept (POC) phase is underway, with no financial commitment made yet.
Under KRA’s strict procurement policy, vendors must demonstrate capability at their own cost.
The AI servers being used in the pilot phase weren’t cheap because each Graphic Processing Unit (GPU) server can cost more than Sh100 million.
“The agreement is clear — we don’t spend a shilling unless we are satisfied that the solution works. We cannot invest in something that costs more than Sh100 million for an experiment. We don’t spend a shilling on proof of concept. If it works, then we talk business,” he says.
The chosen vendor, which already supplies scanning technology to China Customs, agreed to ship two high-powered GPU servers to Kenya to run the demo.
GPU servers, essential for high-speed AI image processing, can cost more than Sh100,000 million each.
“They brought the equipment, set it up and we started training the AI system from scratch by feeding it with thousands of KRA images so it learns to recognise patterns unique to our cargo and declaration habits,” he said.
While language-based AI models like ChatGPT are relatively easy to train, image-based AI for X-ray interpretation is a much tougher challenge.
“This is a totally different beast. It’s not about reading text. It’s about teaching AI to ‘see’ and make intelligent judgements on scanned images,” he said.
Training the system has been underway for about three months.
The early results are promising because the AI is already flagging high-risk shipments and spotting inconsistencies faster than human analysts.
“The system is already making accurate, real-time interceptions.”
KRA expects AI to accurately handle routine, low-risk cargo, allowing human analysts to focus on complex or suspicious cases.
“Just imagine, our officers currently analyse 3,000 images a day. With AI doing 80 per cent of that, humans will only need to look at about 600. That level of efficiency hasn’t been seen since the early 2000s,” Atambo said.
The technology is particularly effective at identifying homogeneous cargo, such as goods imported by Authorised Economic Operators (AEOs), who already benefit from streamlined clearance processes.
“This will free up our human resources to do deeper, more targeted inspections where they matter most,” he added.
KRA is evaluating the proof-of-concept in a live environment, comparing AI-generated reports against human analyses.
This will continue for the next few months, after which a final decision will be made by top management on whether to go ahead with full procurement.
If approved, KRA plans to implement the system by the next financial year. The Authority will also expand the system’s functionality to include risk profiling, historical data mining and automated valuation.
“This is not just about technology, it’s about national revenue and integrity. AI will help us seal loopholes, enhance efficiency and improve service delivery,” Atambo said.
While the promise of AI is immense, the agency is also preparing for the legal and political implications of rolling out the technology.
Public participation and compliance with procurement laws will be critical, especially if the authority proceeds to full acquisition.
“There are already murmurs about public participation and the legalities of introducing new tech. We anticipate some people may try to block the process in court. So we’re preparing to do things by the book,” Atambo said.
There are also concerns – by those who have traditionally exploited gaps – that the technology could disrupt illegal networks in customs controls.
“I’ve already received threats, but I tell people — you can bribe a human, not an AI. AI doesn’t sleep, doesn’t get tired and doesn’t take sides,” he said.
The authority is optimistic that AI will transform cargo handling, particularly at busy borders like Malaba.
This will drastically cut down screening time, especially for empty containers that currently take several minutes due to centralised processing.
“With AI, empty containers can be scanned and cleared in under a second — no stopping, just automatic clearance as they pass through,” he said.
The system also taps into data from KRA’s internal systems including iSCAN, ICMS and the cargo tracking platform RECTS, making its decision-making highly informed and data-driven.
Atambo said KRA and Uganda Revenue Authority have integrated their scanner systems to secure transit cargo along the Northern Corridor.
Once operational, it will automatically compare container images taken at Mombasa Port and Malaba border post.
“If there’s no discrepancy between the two scans, the cargo won’t be stopped at Malaba — it will proceed uninterrupted,” Atambo said.
He believes the result will be smoother trade flows, enhanced security and increased revenue collection.
“AI is the future of customs. It will allow us to process more cargo, more accurately, and in less time — benefiting both the country and legitimate traders,” he concluded
Comments 0
Sign in to join the conversation
Sign In Create AccountNo comments yet. Be the first to share your thoughts!