We've run the operations. That's why we build different things.
Lekke Logic started because the founders kept watching the same project fail at three companies in a row.
The pattern was always the same. A vendor pitched an AI platform. The platform demoed beautifully. Six months in, the project was still in "pilot." Twelve months in, the team had quietly stopped using it. Eighteen months in, someone built a Zapier flow that did 80% of the job for 5% of the cost.
The vendors weren't bad engineers. They were great engineers. They were just building the wrong thing — because they'd never sat in a compliance team at 5pm on a Friday, or watched a sales rep manually triage 200 inbound leads, or seen a support agent burn out on Tier-1 tickets. They were building "platforms" because that's what venture-funded software companies do. The companies buying them didn't need a platform. They needed a system that did a specific job.
So we built one. Then another. Then we did it for a friend's company. Then we did it for the friend's friend. Now it's three years later and that's the company.
We don't sell platforms. We don't sell "frameworks-as-a-service." We don't have a roadmap of features waiting on your data. We build the specific system that does the specific job — in your stack, on your data, with the constraints your business actually has. And we'd rather lose the deal than ship something we wouldn't run ourselves.
Co-founder · Engineering
14 years building production systems at fintechs. Previously led platform engineering at a major SA neobank and shipped the ML stack at a JSE-listed asset manager. Likes: long evals. Dislikes: demos that only work on the demo data.
Co-founder · Operations & Strategy
Ran ops at two scale-ups before this. Lived through the painful side of "AI transformation" projects — twice as the buyer, once as the person who had to clean them up. Now does the audits and the hard conversations.
Senior Engineer
Backend, integrations, the unsexy 60% that makes everything else work. Has opinions about Temporal, Postgres, and people who don't write eval suites.
Plus a rotating bench of 4–5 specialists we've worked with for years. We staff each project with the right people, not whoever happens to be on the bench. We won't grow past 12.
If we recommend a vendor, we've shipped on them. If we recommend an open-source stack, we've run it in production. No referral fees.
Long projects fail more than short ones. We'd rather split a big problem into two 6-week builds with a real release in between.
From day one. No "platform fees." No hostage code. You can fire us next week and the system keeps running.
"Working software" means real users in production. If your project doesn't get there, we haven't done our job.
Sometimes AI isn't the right answer. Sometimes the workflow needs a process change, not a model. We'll say so before you spend a Rand.
We've gone over our own estimate three times in three years — and ate the cost each time. We'd rather lose margin than lose trust.
We're remote-first, but Joburg-based. About half our work is overseas — mostly EU, UK, US East Coast. The other half is South African fintech, SaaS, and media. We're at JoziHub on Tuesdays if you're in town.
30 minutes. No deck. We'll tell you whether to build it — and if so, roughly what it'll take.