Why Opezeni exists
Most AI automates the work. The founder still makes every decision.
Ernest Zimba
Founder — AI/ML engineer
Lusaka, Zambia
I'm self-taught. I started building agentic AI systems in 2026, after spending enough time with AI products to notice what they all had in common: they took over the typing and left the founder holding every judgement call.
Opezeni came out of something smaller. I was building a stress-test agent — a system that pressure-tests a startup idea against its own assumptions before anyone writes code.
Building it raised a question I could not put down: what happens to a SaaS founder's job once AI can make decisions instead of just executing tasks?
That is the whole thesis. Opezeni is a company operating system where specialized agents hold real mandates, act inside guardrails, and escalate only genuine exceptions — so the company keeps moving when the founder is not in the room.
How I think
Operating principles
01
Automate last, not first.
Automation is an amplifier. Point it at a broken process and you get a broken process running faster and costing more. A loop gets automated only after it has been run by hand enough times that the decision rule is boring.
02
Validate before you build.
An idea is an assumption wearing a roadmap. It stays an assumption until a customer pays, switches, or changes behaviour because of it. Every feature has to earn the right to exist before it gets written.
03
Founders should design the future, not operate the present.
Most founder time goes into coordination — approvals, handoffs, status. That work is real, but it is not the work only a founder can do. Opezeni exists to absorb the operating layer so the strategic layer gets the whole calendar.
Track record
What I have actually shipped.
Stated plainly, without rounding up.
Foundation
IBM Professional Certificates — Agentic AI, RAG, Deep Learning
Completed through Coursera. Self-taught before that, and still self-taught after: the certificates formalized what I was already building, they did not replace it.
Shipped
Startup Stress-Test Agent
A LangGraph agent served over FastAPI, with human-in-the-loop checkpoints on every consequential step. Built and deployed. It is the project Opezeni grew out of.
Learning cycles
CogniMerse and AethraSync
Two AI startups I founded before this one. Neither found a commercial market. Both taught me the same lesson from different angles: I was building before I had proof anyone needed it. That lesson is now a rule.
I'm not building another AI tool.
Software companies are entering the same kind of transition as the move from on-premise to cloud. Not a better version of the old shape — a different shape.
Today's AI assists people. Tomorrow's AI will operate systems. The difference is who holds the decision.
The founder's role does not disappear in that world. It moves. From operator to chairman: setting direction, approving exceptions, and leaving the day-to-day to a system built to run it.
I intend to spend a long time on this problem.