Agency in the Age of AI

Most “AI agents” demos are chat with extra steps. Real agency means a system that can take a bounded goal, use tools, and hand work back to a human when it matters. That is the bar we set for African SMEs — and the one we design to.

The junior teammate model

Treat an agent like a capable junior: give it a clear job description, access to the right tools, and a supervisor. Do not give it the keys to payroll on day one. The organisations that win will not be those with the flashiest model — they will be those who define workflows, ownership, and escalation paths before they wire up an API.

What works in production

• Narrow, valuable jobs — brief prep, knowledge retrieval, structured follow-ups — not “run the company.”
• Your data, your infra where risk requires it. Public SaaS is fine for some workflows; it is not fine for all.
• Human control of consequential decisions. Agents propose; people dispose when money, reputation or safety is on the line.
• Measurement. If you cannot name the time or cost the system is supposed to free up, you are buying a toy.

What we told investors

Capital should fund teams that ship systems operators will actually use — not decks that promise transformation. In our own work with founder communities and operating teams, the pattern is consistent: retrieval over retyping, scoring over gut-feel queues, and training that leaves the capability in the organisation when the consultants leave.

Agency in the age of AI is not autonomy theatre. It is designed responsibility: tools that amplify judgment without pretending judgment is optional.