Sihle Letsolo did not set out to build a software product. The founder and Principal Machine Learning Engineer of Olostel AI came to entrepreneurship from machine learning, and what caught his attention was the gap between what artificial intelligence could do on paper and what it was delivering to companies’ bottom lines. Olostel was built to close that gap, and to prove that sophisticated AI work can be engineered in South Africa and sold to the world.
“I didn’t want the company to be limited by the assumption that the most sophisticated technical work has to come from Silicon Valley, Europe or another established technology market,” he says.
The clearest test of that thesis came from Calgary. Olostel had been working through North American manufacturing markets city by city, building prospect lists against its ideal customer profile. The Calgary campaign drew on 576 verified contacts, emailed at a pace of about 40 a day. Fewer than 1% bounced, roughly 4% replied, and three meetings were booked. One was with a chief operating officer whose company became Olostel’s client in a seven-month engagement, with the initial deployment completed in 12 weeks.
What stands out in the pitch is what it left out. The subject line, “Where most COOs leave money on the table”, made no mention of AI. The email introduced the idea of “hidden gaps”, inefficiencies that cost a plant money without showing up clearly in management reports, and closed not with a request for a call but with an offer to send a short summary of the gaps Olostel typically finds. The goal was simply to earn a yes. “The important thing wasn’t that the email was particularly clever,” Letsolo says. What worked was a business question the COO recognised without needing any prior interest in AI.
Winning the contract took more than a reply. The client’s chief executive joined the discussions, and Olostel had to show it understood the operational problem well enough to design around it rather than force an off-the-shelf product onto the plant. Being a South African supplier raised early credibility questions, which Letsolo says faded once conversations turned technical. The work was priced in Canadian dollars, as Olostel generally quotes in the client’s currency, and he rejects the idea that a South African address should mean competing on price.
The deployment itself was substantial: an industrial IoT gateway, integration with SAP Business One and the plant’s manufacturing execution system, a machine-learning service for anomaly detection, alert rules feeding Teams and email, and a Power BI dashboard. Much of the engineering was done from Johannesburg, but Letsolo flew to Calgary, and the time difference meant some fixes waited on plant staff to run physical checks during their shift. “Remote delivery is absolutely possible, but you need the operational discipline to support it,” he says.
Olostel estimates the first-year saving at CAD115,000, or about R1.34 million. Letsolo is careful to separate the evidence: the operational improvements were measured, while the financial figure translates them using the client’s own cost assumptions and has not been independently audited. The downtime numbers are annualised rates, allowing a like-for-like comparison even though the engagement ran for seven months.
He also resists a neat story about overpromising. The plant’s complexity became clearer as the team went deeper, and the response was to break the problem into parts. Staying hands-on in the engineering, he argues, kept the commercial promises and technical reality from drifting apart. “The objective isn’t to prove that your original assumption was right. It’s to get the outcome right.”
Today Olostel employs 11 people and runs more than 60 AI agents across its own operations and client environments, handling tasks such as administration, invoice collection and content writing. It has seven active retainer clients alongside project work of three to six months, and annual revenue above R10 million. The privately held company is in talks with prospective investors and declines to disclose profitability, its funding position or the share of revenue earned offshore.
For Letsolo, the hard parts of exporting technical services are rarely technical: credibility with clients who have never hired a South African tech firm, time zones, payment terms, procurement, data access and the occasional need to stand on a plant floor. Over the next five years he expects his time to tilt further towards strategy, partnerships and growth, but he has no plan to leave engineering behind. Technical credibility, in his view, is part of the founder’s edge.