The Kenya Revenue Authority has recovered Ksh 4 billion in previously undisclosed tax liabilities through an artificial intelligence-driven compliance engine deployed at scale in the financial year ending June 2026, Commissioner-General Humphrey Wattanga announced at the authority’s performance review briefing in Nairobi this week. The figure, which KRA describes as a direct attribution to the AI system rather than general enforcement activity, represents a significant early return on a Ksh 780 million investment in the technology made under the 2024 Medium-Term Revenue Strategy.
The AI system, developed in partnership with a consortium of local and international technology firms under a competitive tender, operates across five analysis modules: VAT gap detection, income underreporting in the SME sector, digital economy revenue tracking, real estate income reconciliation, and customs valuation fraud identification. Each module trains on a combination of KRA’s own historical declaration data, third-party financial flows, and cross-agency data sharing with institutions including the Land Registry, the Central Bank’s Financial Intelligence Unit, and the Insurance Regulatory Authority.
How the System Finds Hidden Money
The VAT module has proven the most immediately productive. By comparing declared turnover against M-Pesa Lipa Na M-Pesa transaction flows — accessible under a data-sharing agreement with Safaricom signed in 2024 — the system identifies businesses whose declared VAT-able revenues are statistically inconsistent with their actual customer transaction volumes. Wattanga said the module had generated 4,200 audit flags in its first operational year, of which 1,840 were escalated to compliance officers and 620 resulted in assessed additional liabilities.
“A trader who declares Ksh 800,000 in monthly turnover but processes Ksh 3.2 million through their till M-Pesa is flagged automatically. The compliance officer receives a pre-populated audit file with three years of transaction history,” said Wattanga. “The human’s job is no longer to find the discrepancy. It is to assess it and pursue it.”
The digital economy module targets a category of taxpayer that KRA acknowledges it has historically struggled to capture: freelancers, content creators, and micro-entrepreneurs earning through international platforms including Fiverr, Upwork, YouTube, and various e-commerce aggregators. By monitoring foreign exchange inflows through the CBK’s reporting system and cross-matching them against income tax returns, the system identified approximately 28,000 individuals receiving regular foreign currency income but filing minimal or no income tax returns.
Real Estate and the Wealthy
The real estate reconciliation module has attracted the most political attention. By cross-referencing Land Registry transfer data against market valuations and mortgage drawdowns visible through banking system data, the AI flags transactions where declared sale prices are implausibly low relative to known market rates — a well-documented form of both tax evasion and illicit wealth transfer in Kenya’s property market.
KRA declined to name individuals identified through the real estate module, citing ongoing legal proceedings, but Wattanga confirmed that 14 high-value assessments totalling Ksh 1.1 billion had been raised against property transactions in Nairobi, Mombasa, and Naivasha, with collections of Ksh 340 million already secured and the remainder under dispute or enforcement.
The tool carries clear political symbolism in the context of Kenya’s IMF programme. Treasury Cabinet Secretary John Mbadi has repeatedly faced criticism from Gen Z-aligned civil society groups who argue that austerity measures disproportionately burden ordinary Kenyans while wealthy elites evade liability. A KRA AI system that visibly pursues elite tax evasion rather than squeezing PAYE workers offers the government a politically useful counter-narrative heading into 2027.
Expansion Plans
KRA intends to extend the AI system’s scope in the 2026/27 financial year to include withholding tax compliance in the professional services sector and excise duty monitoring for alcohol and tobacco manufacturers. The authority’s medium-term target, embedded in the National Tax Policy approved in 2025, is to raise the country’s tax-to-GDP ratio from its current 14.7 per cent to 18 per cent by 2030. The AI compliance programme is projected to contribute an incremental Ksh 25 billion annually by the time all modules reach full maturity — a target that independent economists describe as ambitious but not implausible given the evidence of the first year’s results.
0 comments