Market and Systems Interventions

The skills landscape: AI has restructured the problem

The talent and skills ecosystem challenge requires solutions at both ecosystem and systemic levels, operating on different time horizons. The landscape has changed materially since 2022 - most significantly through AI, which has restructured what skills are scarce, what can be taught, and how learning can be delivered. The substantive treatment of the engineering-to-management-scarcity shift, anchored in Acemoglu and Restrepo's framework on automation and labour markets, sits in Leadership and Human Capital. The implication for ecosystem-level intervention is that the skills landscape now divides cleanly between displacement-effect categories (where AI substitutes for labour) and productivity-complement categories (where AI raises the value of human capability). The talent infrastructure being built across the continent must be calibrated to this distinction or it will produce capability for the labour market of five years ago.

The 2021 jobtech map, developed alongside the Jobtech Alliance - an ecosystem-building initiative steered by Mercy Corps and BFA Global - identified 323 active innovators in the African jobtech space. The Jobtech Alliance taxonomy organises these into five categories: platforms for offline work, platforms for digitally-delivered work, digital services for micro-enterprises, technology-enabled skilling, and digital tools for worker enablement. Enabler-type companies - startups building infrastructure for other startups - are contributing a multiplier effect that extends beyond direct employment creation.

The structural significance of this digital-labour-market layer is named in the contemporary economics of the African digital economy. Hjort and Poulsen's American Economic Review paper "The Arrival of Fast Internet and Employment in Africa" - drawing on submarine fibre-optic cable arrival across multiple African countries - established the foundational empirical pattern: digital infrastructure produces measurable, large-scale labour market effects in African economies, with employment gains concentrated in higher-skill formal-sector roles and the effects compounding over time as the infrastructure becomes denser. The implication for the jobtech landscape is direct: the 323 innovators Mercy Corps and the Jobtech Alliance map are not a peripheral category. They are the operational infrastructure through which the digital labour-market effects Hjort and Poulsen documented now propagate at scale.

Matching platforms include Afringa, an online recruitment platform active in more than 40 African countries that uses video applications and machine learning to address two structural problems in African recruitment: credential fraud and skills-qualification mismatch. Talent-networked models include Venture for Africa, which provides talent with a landing pad in the African startup ecosystem through part-time, trial-to-hire immersion engagements. Training providers including Moringa School,Semicolon, and Gebeya - the substantive treatment of which as a labour platform sits in Platform Models - have sustained demand by aligning curricula directly with employer needs rather than academic programme structures.

The skills conversation in 2022 was dominated by the coding shortage. AI has partially - and in some domains, substantially - changed this calculus. The skills AI cannot yet replicate - complex reasoning, empathy, contextual judgment, cross-cultural communication, ethical decision-making - are precisely those that African education systems have historically underinvested in. The irony is direct: the skills Africa's education system treats as peripheral are the skills the AI economy values most.

Accelerators and hubs should be harvesting significant labour market data. Many of these programmes are funded by public or donor money - there should be requirements to collect, report, and share relevant labour market datasets as a condition of funding. The substantive treatment of why data architecture in the African scaling ecosystem requires institutional design rather than goodwill - anchored in Hess-Ostrom on knowledge as a commons and Benkler on commons-based knowledge production - sits in Data, Insights and Knowledge. The implication for talent intelligence specifically is that no such conversation is happening systematically within African ecosystems about designing collective intelligence approaches to the talent problem.

AI-assisted labour market analytics now make this more achievable than it was in 2022: platforms that aggregate job posting data, skills demand signals, and graduate outcome data can generate labour market intelligence at a granularity and frequency that traditional survey-based approaches cannot match. The tools exist. The institutional will to share the data and act on the intelligence is what is missing.

Mental health support for African founders deserves elevation from a peripheral consideration to a core design requirement of scaling support programmes. The substantive treatment of the entrepreneurial mental health evidence base - anchored in Freeman, Staudenmaier, Zisser and Andresen's Small Business Economics finding that 72 percent of entrepreneurs experience mental health differences directly or indirectly, and extending to Stephan and Wiklund-Hatak-Patzelt-Shepherd on entrepreneurial wellbeing dynamics - sits in The 15 Acceleration Goals Goal 12. The implication for African-specific support design is that the African context - where founders carry the weight of community expectations, family obligations, and the structural volatility of the operating environment - compounds these pressures in ways the existing literature, drawn almost entirely from Western entrepreneurial contexts, does not fully capture.

"My entire village depends on me succeeding. In Africa, you have to take care of everybody. That's where the resilience comes from." - interviewee

"The thing I keep telling people is that if you're a little bit suicidal, a little bit bipolar, emotionally, physically, spiritually, and destroyed all of your relationships, and gone completely bankrupt: Welcome to your first five seconds of being an entrepreneur. It's not this beautiful, lovely rose-coloured thing." - interviewee

A framework for action, drawing on the work of Marcel Muenster's Gritti Fund - described as the first investment fund to incorporate founder mental and physical performance as an explicit investment criterion, with a specific focus on healthcare investments in the Middle East and Africa - includes three elements: destigmatisation, requiring investors to lead by example and demonstrate that vulnerability about mental health challenges is acceptable; wellbeing resources, integrating mental health professionals into the organisational ecosystem of investment funds rather than treating them as external referrals; and structural investment, making a defined fraction of investment explicitly available for founder wellbeing rather than treating it as a cost to be minimised.

"Managing mental health has to start from the donor, the funder. One of the metrics of their return on investment should also be the well-being of the beneficiary." - interviewee

Smart investors dedicate significant time to the talent challenge - from seed to Series B - by making the right connections, leveraging internal value creation teams, and participating in recruitment at C-suite level. The talent pipeline is not a collective action problem that individual investors can solve in isolation.

The structural mechanic underlying the difficulty is named in Marcel Fafchamps' research on African business networks. Fafchamps' Market Institutions in Sub-Saharan Africa - drawing on extensive empirical research across multiple African economies - established that African business networks operate substantially through interpersonal trust mechanisms rather than formal institutional channels: information flows, capital flows, and crucially, talent flows track relational ties more than they track market signals. The implication for ecosystem-level talent infrastructure is direct. Syndicating interest around the pipeline - sharing data, coordinating investments in training infrastructure, building shared recruitment intelligence - would return better results than current zero-sum approaches in which each investor builds proprietary talent networks. But because African business networks operate on the trust-mechanism architecture Fafchamps describes, the syndication infrastructure has to be built explicitly as institutional design rather than emerging from market behaviour. The default outcome of leaving network architecture to evolve organically is the proprietary-network equilibrium that sustains the gap.

Diaspora engagement remains a significant and underutilised opportunity. The substantive treatment of the diaspora-knowledge-network dynamic - anchored in Saxenian's brain-circulation framework - sits in Ecosystem Characteristics and Founders and Leadership Teams. Ethiopia's Diaspora Tech Homecoming initiative, part of the broader Great Ethiopian Homecoming launched by the Prime Minister in December 2021, demonstrated what deliberately designed diaspora engagement can look like at government level. Full return is unlikely for most diaspora - but partial engagement through angel syndicates, remote mentoring, internship sponsorship, and diplomatic advocacy creates genuine value without requiring the career disruption that full repatriation involves.

AI-assisted assessment is changing how skills can be verified. Rather than relying on credentials from institutions whose quality is difficult to calibrate across 54 countries, ventures can now use skills-based assessment tools that test competency directly. The HNG Internship - founded by Hotels.ng's Mark Essien, running since 2016, and now one of the largest remote tech internship programmes in Africa - takes candidates through multiple rounds of task-based, impartial assessment. Finalists demonstrate competency directly rather than through institutional affiliation. The model points toward the direction of travel: assess what people can do, not where they studied.

A survey of colleges and universities in sub-Saharan Africa found that only 7 percent have an entrepreneurship centre, 28 percent offer courses specialising in entrepreneurship, and 10 percent offer a course in innovation and technology. The root cause issues begin earlier: formal school education systems remain largely based on models designed to produce labour for the first industrial revolution. Many African universities still operate large lecture halls that reward the ability to remember and repeat information - structurally incompatible with the skills the AI economy demands.

The Hjort-Poulsen finding that digital infrastructure produces measurable employment effects has a less-cited corollary that applies directly here: the largest gains concentrate in higher-skill formal-sector occupations and depend on the availability of tertiary-educated labour to capture them. Internet infrastructure produces broad-based labour-market gains in their data, but the formal-sector job creation that drives sustained productivity improvement depends on the human-capital complement being there to exploit it. The current African education-systems mismatch is therefore not just a labour-market problem but an infrastructure-effectiveness problem: the digital infrastructure being built across the continent will produce the labour-market gains African economies need only if the human-capital complement is built in parallel.

The ecosystem is building complementary infrastructure. Andela has pivoted from primarily training junior software engineers to a broader talent marketplace model - the substantive treatment of this pivot as a strategic-decision case sits in The Scaling Decision Log Decision 4. AltSchool Africa and Refactory in Uganda are developing models that align training directly with employer demand rather than academic curricula. The non-profit model proposed by Future Africa - bringing together AltSchool Africa, Andela, Gebeya, and others to work with established universities on an evolving curriculum - has real merit. Its design and implementation must embrace learner-centred approaches explicitly oriented toward the AI-augmented workplace, not toward the skills profile of the economy that existed when those universities were established.