Zaajira: Developing an AI Talent Platform

In much of Africa, recruitment volume is not the constraint. Employers advertising entry-level and early-career roles routinely receive far more applications than they can review, while still struggling to fill the role — because the applications that arrive rarely evidence the specific competencies the requisition needs. The result is a market where both sides work hard and neither gets what it wants: recruiters burn their week on first-pass review, and candidates are rejected without ever learning what was missing.

Zaajira was designed around that mismatch. Instead of treating sourcing as the product, it treats qualification as the product: rank candidates against the requisition using structured competency signals, screen them asynchronously so recruiter time is spent on shortlists, and route candidates with identified gaps into targeted learning rather than into a rejection email.

A requisition is captured as structured requirements rather than free text, which is what makes ranking meaningful. Candidate profiles are scored against those requirements, and shortlisted candidates move into an async screening stage — audio and video screeners with calibrated rubrics, so two recruiters scoring the same answer arrive at comparable results. Interview kits are configurable per role family, which keeps the process consistent across countries and hiring managers.

The skilling layer is built on LearnWorlds and mapped to in-demand role competencies. When a candidate is flagged for a specific gap, the platform can enrol them in the matching pathway in the same session. That turns a rejection into a re-engagement pipeline: the employer keeps a warm pool of candidates who are actively closing the exact gap that blocked them, and the candidate gets actionable feedback instead of silence.

Recruiters work from dashboards covering pipeline health, time-to-shortlist, source effectiveness and funnel drop-off. Employer admins get cohort reporting that connects hiring outcomes to skilling participation — the reporting that makes long-term workforce development defensible internally rather than aspirational.

Building for multiple African markets imposes constraints that are easy to miss from a Gulf or European office. Connectivity is intermittent and metered, so the candidate experience was tuned for low bandwidth: compressed media capture, resumable uploads, and screens that degrade rather than break when a request fails. Mobile is the primary and often only device, which shaped navigation, form length and file handling.

Identity and document handling had to accommodate differing national ID standards and right-to-work verification requirements across markets, with documents stored and accessed under least-privilege rules. Multi-language support was treated as a data-model concern, not a translation layer bolted on at the end.

The structural logic behind Zaajira is not Africa-specific. Any market with high applicant volume and a competency gap — including Gulf employers hiring at scale under national workforce programmes — benefits from the same three moves: structure the requisition, screen asynchronously with calibrated scoring, and attach learning to rejection. The World Bank and the ILO both document the scale of the youth labour-market mismatch that makes embedded skilling a hiring lever rather than a CSR line item.

If you are building a talent platform, the transferable lesson is to keep one shared record across hiring and learning. The moment those live in separate systems, long-term talent development stops being measurable — and anything unmeasurable eventually stops being funded.