London-listed accounting software firm Sage has launched an AI platform it said will ease the administrative burden on small business owners across Africa and the Middle East. But it enters a market flooded with similar promises – and cautious customers who cannot afford to get their finances wrong.
The company unveiled Sage AI in Johannesburg this week, embedding artificial intelligence into its finance, payroll and HR products. The timing is deliberate. South African small businesses are under pressure: rising costs, skills shortages and compliance demands are stretching owners thin. Sage is betting AI can fix that. Its rivals, including Microsoft, Xero and QuickBooks, are making the same bet.
Jordaan Burger, MD for Sage Africa and the Middle East, acknowledged the noise. “There are very few platforms these days that will not tell you they are AI-enabled,” he said. “I want to challenge you to go deeper – where do you actually apply it, what is the benefit, and how can you trust it?”
That last word – trust – is where Sage is planting its flag. In the finance and payroll world, it is not a marketing term but a practical necessity. Get payroll wrong and employees don’t get paid correctly. Get tax submissions wrong and Sars comes knocking. For the small business owners Sage is targeting, those are not theoretical risks.
“Getting the numbers mostly right is not right enough,” Burger said. “AI should remove risk from your systems, not introduce more.”
The platform runs on what Sage said are roughly 40 000 domain-specific AI models trained on financial, payroll and HR data. The company claims the platform has saved customers 50 million hours globally and that one of its AI agents can cut the month-end close cycle by around 90%.
Can it be trusted?
But if Sage AI makes a mistake, who takes the fall?
“In a perfect world, I want to believe that our testing is sufficient so that our AI doesn’t make a mistake,” Burger said, pointing to beta-testing and customer feedback loops as safeguards.
On whether AI would replace accountants, Burger was more definitive: “Absolutely not! We want to keep humans in control. What we are trying to give back is time – to focus on what really matters.”
That human-in-the-loop principle is built into the platform’s architecture. AI agents can flag anomalies and recommend actions but cannot execute them without human sign-off.
Read: The AI jobs reckoning is here
Yolandi Esterhuizen, Sage’s product compliance director, made that concrete. If an AI agent detects a salary variance, it logs the change and explains its reasoning. Nothing moves without approval. “No financial adjustment will automatically be made,” she said. “All actions are auditable.”
Esterhuizen said compliance is designed into the product from day one. “Regulators like Sars are increasingly focusing on automated risk profiling, data matching and anomaly detection,” she said. “That is why we are extremely deliberate about how we apply AI within our systems.”
Sophia Adhami, Sage’s senior director of product performance, outlined from London how the platform handles data. Sage said models are trained on anonymised, customer-specific data that is never shared across users or used to train external models. The platform adheres to global standards including the EU AI Act, and Sage says it is sponsoring research to limit AI hallucinations – where AI systems produce confident but incorrect outputs.
Adhami also introduced an AI Trust Label, an in-product disclosure panel she compared to a food nutrition label. Users can click it to see what a feature does, what data it draws on and which compliance standards apply.
‘What really matters’
The launch also carries a policy dimension. The government recently raised the VAT registration threshold from R1-million to R2.5-million. Burger warned against reading that as a reason to delay technology investment. “Compliance is the absolute minimum outcome,” he said. “Technology should be used to get ahead and to beat your competitors.”
Read: AI is breaking the link between university degrees and employment
For businesses across Africa still operating with limited digital infrastructure, Adhami said the strategy is to embed AI into existing products so that automating even a single workflow – she pointed to e-invoicing – delivers immediate value without requiring full digital transformation upfront. “What really matters is getting paid, and getting paid as quickly as you can,” she said. – © 2026 NewsCentral Media
Get breaking news from TechCentral on WhatsApp. Sign up here.