Hanademi

When AI helps entrepreneurs, and when it hurts

35 slides · $4.57 · 29 min · 2 minutes ago language
ENES
theme
LightDark
view
DeckTableTalk
brand
HanademiPlatzi
When AI helps entrepreneurs,and when it hurtsMade for Freddy Vega, by Hanademi
Cuándo la IA ayuda aemprender y cuándo perjudicaMade for Freddy Vega, by Hanademi
The same AI mentor helped the strongestand hurt the weakestEstimated business-performance effect in a randomized study of 640 Kenyan entrepreneurs using a GPT-4mentor through WhatsApp.Made for Freddy Vega, by HanademiSources: Otis, N., Clarke, R., Delecourt, S., Holtz, D., & Koning, R. (2024). The uneven impact of generative AI on entrepreneurialperformance. Harvard Business School Working Paper.Unitestimated effectHigher baseline performance15%Lower baseline performance-8%
A Kenyan experiment gave 640 entrepreneurs the same AI business mentor. Stronger participants improved about 15%, while weaker participants declined about 8%. The product was identical, but the users were not. Programs that ignore capability can widen the gap they hope to close.
El mismo mentor de IA ayudó a los másfuertes y perjudicó a los más débilesEfecto estimado sobre el desempeño empresarial en un estudio aleatorizado con 640 emprendedoreskenianos que usaron un mentor GPT-4 por WhatsApp.Made for Freddy Vega, by HanademiFuentes: Otis, N., Clarke, R., Delecourt, S., Holtz, D., & Koning, R. (2024). The uneven impact of generative AI on entrepreneurialperformance. Harvard Business School Working Paper.Unidadefecto estimadoMayor desempeño inicial15 %Menor desempeño inicial-8 %
Un experimento en Kenia dio el mismo mentor empresarial de IA a 640 emprendedores. Los participantes más fuertes mejoraron cerca de 15%, mientras los más débiles cayeron cerca de 8%. El producto era idéntico, pero los usuarios no. Los programas que ignoran las capacidades pueden ampliar la brecha que buscan cerrar.
The terms behind the argumentMade for Freddy Vega, by HanademiGenerative AIAI that creates new text, images, audio or software in response to instructions.Informal employmentWork lacking some legal registration, social protection, benefits or formalcontractual safeguards.CounterfactualThe best estimate of what would have happened without the intervention.
3 technical terms carry much of the argument.
Los términos detrás del argumentoMade for Freddy Vega, by HanademiIA generativaIA que crea texto, imágenes, audio o software nuevos en respuesta a instrucciones.Empleo informalTrabajo que carece de algún registro legal, protección social, prestaciones osalvaguardias contractuales formales.ContrafactualLa mejor estimación de lo que habría ocurrido sin la intervención.
3 términos técnicos sostienen buena parte del argumento.
Sub-Saharan self-employment barely fellbelow 78%Self-employment as a share of total employment, modeled ILO estimates for Sub-Saharan Africa,1991-2025. The series was not independently verified at build time.Made for Freddy Vega, by HanademiSources: World Bank. (2024). Self-employed, total (% of total employment), modeled ILO estimate. World Development Indicators.Unit% of employment78%80%1991199620012006201120162021The modeled rate peaked at 81.2%in 1995.It remains 77.7% in 2025.80.8%
Self-employment has remained the dominant employment form across Sub-Saharan Africa for decades. The modeled rate eased from above 81% to 77.7% in 2025. That is a huge potential user base for entrepreneurial support. It is not evidence that every livelihood business is ready to scale.
El autoempleo subsahariano apenas cayópor debajo de 78%Autoempleo como proporción del empleo total, estimaciones modeladas de la OIT para África subsahariana,1991-2025. La serie no se verificó de forma independiente al construir.Made for Freddy Vega, by HanademiFuentes: World Bank. (2024). Self-employed, total (% of total employment), modeled ILO estimate. World Development Indicators.Unidad% del empleo78 %80 %1991199620012006201120162021La tasa modelada alcanzó un picode 81,2% en 1995.Permanece en 77,7% en 2025.80,8 %
El autoempleo ha sido la forma dominante de empleo en África subsahariana durante décadas. La tasa modelada bajó de más de 81% a 77,7% en 2025. Es una enorme base potencial para apoyo empresarial. No prueba que todo negocio de subsistencia esté listo para escalar.
Latin American self-employment settlednear 36%Self-employment as a share of total employment in Latin America and the Caribbean, modeled ILOestimates, 1991-2025.Made for Freddy Vega, by HanademiSources: World Bank. (2024). Self-employed, total (% of total employment), modeled ILO estimate. World Development Indicators.;World Bank. (n.d.). Self-employed, total (% of total employment) (modeled ILO estimate) [SL.EMP.SELF.ZS]. World Bank Open Data.Unit% of employment34%36%38%40%1991199620012006201120162021The rate peaked at 39.6% in1993.It ends at 35.6% in 2025.
Latin America and the Caribbean followed a lower and flatter self-employment path. The modeled rate declined from about 39% in the early 1990s to 35.6% in 2025. That baseline is less than half Sub-Saharan Africa's modeled level. A single global entrepreneur program would therefore begin with very different user populations.
El autoempleo latinoamericano se estabilizócerca de 36%Autoempleo como proporción del empleo total en América Latina y el Caribe, estimaciones modeladas de laOIT, 1991-2025.Made for Freddy Vega, by HanademiFuentes: World Bank. (2024). Self-employed, total (% of total employment), modeled ILO estimate. World Development Indicators.; WorldBank. (n.d.). Self-employed, total (% of total employment) (modeled ILO estimate) [SL.EMP.SELF.ZS]. World Bank Open Data.Unidad% del empleo34 %36 %38 %40 %1991199620012006201120162021La tasa alcanzó un pico de39,6% en 1993.Termina en 35,6% en 2025.
América Latina y el Caribe siguió una trayectoria de autoempleo más baja y estable. La tasa modelada bajó de cerca de 39% a inicios de los años noventa a 35,6% en 2025. Esa base es menos de la mitad del nivel modelado de África subsahariana. Un programa empresarial global comenzaría con poblaciones muy diferentes.
ILO estimates put 86% of Africanemployment outside formalityInformal employment as a share of employment by region, ILO statistical estimates published in 2018. Thefigures were not live-verified at build time.Made for Freddy Vega, by HanademiSources: International Labour Organization. (2018). Women and men in the informal economy: A statistical picture (3rd ed.). ILO.Unit% of employment85.8%Africa68.6%Arab States68.2%Asia-Pacific40%Americas25.1%Europe and Central Asia
The entrepreneurial base is not only large. It is overwhelmingly informal. The ILO estimate reaches 85.8% in Africa, more than triple the European and Central Asian rate. Identity checks, formal accounts and credit records can therefore filter out intended users.
La OIT estima que 86% del empleo africanoestá fuera de la formalidadEmpleo informal como proporción del empleo por región, estimaciones estadísticas de la OIT publicadas en2018. Las cifras no se verificaron en vivo al construir.Made for Freddy Vega, by HanademiFuentes: International Labour Organization. (2018). Women and men in the informal economy: A statistical picture (3rd ed.). ILO.Unidad% del empleo85,8 %África68,6 %Estados árabes68,2 %Asia-Pacífico40 %Américas25,1 %Europa y Asia central
La base emprendedora no solo es grande. Es abrumadoramente informal. La estimación de la OIT llega a 85,8% en África, más del triple de la tasa de Europa y Asia central. Los requisitos de identidad, contabilidad formal e historial crediticio pueden excluir a los usuarios previstos.
More than 9 crore establishmentswere counted, and informality doesnot mean offline.The same release reported more than 15 crore workers. Digital participation is achannel indicator, not an income outcome.Sources: mospi.gov.in.
India counted more than 9 crore unincorporated establishments in early 2026. Eight in ten were reported as online and cashless. That creates a vast digital distribution opportunity. It does not establish whether owners have the skills, trust or time to use AI well.
Se contaron más de 9 crore deestablecimientos, y la informalidad nosignifica desconexión.El mismo informe reportó más de 15 crore de trabajadores. La participación digital esun indicador de canal, no un resultado de ingresos.Fuentes: mospi.gov.in.
India contó más de 9 crore de establecimientos no incorporados a inicios de 2026. Ocho de cada diez fueron reportados como digitales y sin efectivo. Eso crea una enorme oportunidad de distribución digital. No establece si sus propietarios tienen las habilidades, confianza o tiempo para usar bien la IA.
The $5.2 trillion gap excludes mostinformal businesses.The estimate concerns formal micro, small and medium enterprises and should not beapplied directly to informal firms.Sources: International Finance Corporation. (2017). MSME finance gap: Assessment of the shortfalls and opportunities in financing micro, small, andmedium enterprises in emerging markets. IFC.
Capital is already scarce before AI enters the picture. IFC estimated a $5.2 trillion annual financing gap for formal small and medium businesses in developing economies. Most informal enterprises were outside that estimate. Advice without capital may improve plans that owners still cannot finance.
La brecha de 5,2 billones excluye a lamayoría de negocios informales.La estimación corresponde a micro, pequeñas y medianas empresas formales y no debeaplicarse directamente a empresas informales.Fuentes: International Finance Corporation. (2017). MSME finance gap: Assessment of the shortfalls and opportunities in financing micro, small, andmedium enterprises in emerging markets. IFC.
El capital ya es escaso antes de introducir IA. La IFC estimó una brecha anual de 5,2 billones de dólares para pequeñas y medianas empresas formales en economías en desarrollo. La mayoría de empresas informales quedó fuera. La asesoría sin capital puede mejorar planes que sus dueños todavía no pueden financiar.
Nigeria allocated $34 million and trackedapplicants for 5 years.The competition tested business grants and selection, not AI-enabled entrepreneursupport.Sources: pdf.usaid.gov.
Nigeria's randomized business-plan competition treated capital and measurement as seriously as recruitment. It allocated $34 million, roughly $50,000 per winner, and followed applicants for five years. That is the standard of evidence an income claim demands. AI programs should be judged against outcomes with similar patience.
Nigeria asignó 34 millones de dólares ysiguió solicitantes durante 5 años.El concurso evaluó subvenciones empresariales y selección, no apoyo empresarialmediante IA.Fuentes: pdf.usaid.gov.
El concurso aleatorizado de planes de negocio de Nigeria trató el capital y la medición con la misma seriedad que el reclutamiento. Asignó 34 millones de dólares, cerca de 50.000 por ganador, y siguió a los solicitantes durante cinco años. Ese es el nivel de evidencia que exige una afirmación sobre ingresos. Los programas de IA deben evaluarse con una paciencia similar.
AI raised productivity 14% to 40% onbounded workReported productivity improvements from workplace experiments with different tasks and populations,2023 studies.Made for Freddy Vega, by HanademiSources: Brynjolfsson, E., Li, D., & Raymond, L. R. (2023). Generative AI at work. NBER Working Paper No. 31161.; Noy, S.,& Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science.;Dell'Acqua, F., et al. (2023). Navigating the jagged technological frontier. Harvard Business School Working Paper 24-013.Unitreported improvementSupport agents14%Largest reported effect40%2.9x
Several workplace experiments found meaningful productivity gains. One study of 5,179 support agents reported 14% more issues resolved per hour, while the largest reported improvement across three experiments reached about 40%. These were bounded tasks with observable outputs. Informal entrepreneurship involves messier decisions and weaker feedback.
La IA elevó la productividad entre 14% y40% en tareas delimitadasMejoras de productividad reportadas en experimentos laborales con distintas tareas y poblaciones, estudiosde 2023.Made for Freddy Vega, by HanademiFuentes: Brynjolfsson, E., Li, D., & Raymond, L. R. (2023). Generative AI at work. NBER Working Paper No. 31161.; Noy, S., &Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science.; Dell'Acqua, F., etal. (2023). Navigating the jagged technological frontier. Harvard Business School Working Paper 24-013.Unidadmejora reportadaAgentes de atención14 %Mayor efecto reportado40 %2,9x
Varios experimentos laborales encontraron mejoras importantes de productividad. Un estudio con 5.179 agentes reportó 14% más casos resueltos por hora, mientras la mayor mejora entre tres experimentos llegó a cerca de 40%. Eran tareas delimitadas con resultados observables. El emprendimiento informal implica decisiones más ambiguas y menos retroalimentación.
Anthropic estimated an 80% reduction inindividual task time, without provinghigher worker income.The estimate concerns individual task time and does not measure earnings, businessprofit or survival.Sources: anthropic.com.
Anthropic analyzed 100,000 Claude conversations and estimated an 80% reduction in individual task time. The counterfactual task averaged roughly 90 minutes. That is a striking capability signal. It remains a company-produced task estimate, not evidence that entrepreneurs earned more.
Anthropic estimó una reducción del 80% enel tiempo de tareas individuales, sin probarmayores ingresos laborales.La estimación corresponde al tiempo de tareas individuales y no mide ingresos,utilidades empresariales ni supervivencia.Fuentes: anthropic.com.
Anthropic analizó 100.000 conversaciones con Claude y estimó una reducción del 80% en el tiempo de tareas individuales. La tarea contrafactual promediaba cerca de 90 minutos. Es una señal llamativa de capacidad. Sigue siendo una estimación empresarial de tareas, no evidencia de que los emprendedores ganaron más.
Across 7,000 workplaces, chatbots did notsignificantly raise earnings or hours.The supplied citation is secondary reporting of the study, so the result should beinterpreted cautiously.Sources: fortune.com.
Task productivity is not the same as economic impact. A study covering 7,000 workplaces found no significant chatbot effect on earnings or recorded hours in any occupation. That does not prove AI never affects income. It shows that time savings do not automatically flow into workers' pay.
En 7.000 lugares, los chatbots no elevaronsignificativamente ingresos ni horas.La cita corresponde a cobertura secundaria del estudio, por lo que el resultado debeinterpretarse con cautela.Fuentes: fortune.com.
La productividad de tareas no equivale a impacto económico. Un estudio de 7.000 lugares de trabajo no encontró efectos significativos de los chatbots sobre ingresos ni horas registradas en ninguna ocupación. Eso no prueba que la IA nunca afecte los ingresos. Muestra que el tiempo ahorrado no se convierte automáticamente en remuneración.
A screen is not an outcomeA woman works at a computer in an office setting.Sources: fortune.com.
Visible technology use must not be mistaken for higher earnings.
Una pantalla no es un resultadoUna mujer trabaja con una computadora en una oficina.Fuentes: fortune.com.
El uso visible de tecnología no debe confundirse con mayores ingresos.
AI users were 19 percentage points lesslikely to answer correctly, a change of -19percentage points.The finding concerns a specific consultant task and does not imply a -19-point penalty forevery AI use.Sources: Dell'Acqua, F., et al. (2023). Navigating the jagged technological frontier. Harvard Business School Working Paper 24-013.; What is the jaggedAI frontier?.
Speed is not the only outcome that matters. Consultants using AI on a task outside its reliable frontier were 19 percentage points less likely to produce the correct answer, a change of -19 points. Entrepreneurs often cannot independently verify legal, financial or agricultural advice. A confident error can cost more than the time AI saves.
Los usuarios de IA tuvieron 19 puntosporcentuales menos de probabilidad deresponder correctamente, un cambio de-19 puntos porcentuales.El resultado corresponde a una tarea específica de consultoría y no implica unapenalización de -19 puntos para todo uso de IA.Fuentes: Dell'Acqua, F., et al. (2023). Navigating the jagged technological frontier. Harvard Business School Working Paper 24-013.; What is the jaggedAI frontier?.
La velocidad no es el único resultado importante. Los consultores que usaron IA fuera de su frontera confiable tuvieron 19 puntos porcentuales menos de probabilidad de responder correctamente, un cambio de -19 puntos. Los emprendedores muchas veces no pueden verificar asesoría legal, financiera o agrícola. Un error convincente puede costar más que el tiempo ahorrado.
Hustle Academy can recruit at scale beforeprofits are knownAbsa Bank Kenya and Google targeted 3,000 small businesses over 12 months; more than 600 had enrolledacross three initial cohorts. Togo mobile subscriptions show channel growth.Made for Freddy Vega, by HanademiSources: tech-ish.com.; Google. (2024). Hustle Academy program materials. Google.; World Bank. (n.d.). Mobilecellular subscriptions (per 100 people) [IT.CEL.SETS.P2]. World Bank Open Data.Unitbusinesses and subscriptionsProgram recruitment12-month target · 3,000Initial enrollmentTogo mobile channel1.080.820002024
The Kenya program targeted 3,000 small businesses over 12 months and enrolled more than 600 in its first three cohorts. Mobile infrastructure makes broad recruitment increasingly plausible. Public program materials do not establish effects on profit, revenue or survival. Reach and business health remain separate questions.
Hustle Academy puede reclutar a escalaantes de conocer las utilidadesAbsa Bank Kenya y Google fijaron una meta de 3.000 pequeñas empresas durante 12 meses; más de 600 sehabían inscrito en tres cohortes iniciales. Las suscripciones móviles de Togo muestran el crecimiento delcanal.Made for Freddy Vega, by HanademiFuentes: tech-ish.com.; Google. (2024). Hustle Academy program materials. Google.; World Bank. (n.d.). Mobile cellularsubscriptions (per 100 people) [IT.CEL.SETS.P2]. World Bank Open Data.Unidadempresas y suscripcionesReclutamiento del programaMeta de 12 meses · 3.000InscripciónCanal móvil de Togo1,080,820002024
El programa de Kenia fijó una meta de 3.000 pequeñas empresas durante 12 meses e inscribió a más de 600 en sus primeras tres cohortes. La infraestructura móvil hace más viable el reclutamiento amplio. Los materiales públicos del programa no establecen efectos sobre utilidades, ingresos o supervivencia. Alcance y salud empresarial siguen siendo preguntas distintas.
FarmerChat passed 1 million reportedusers before profit impact was knownOrganization-reported FarmerChat reach from a 2024 figure and a later update. The latest value is a lowerbound.Made for Freddy Vega, by HanademiSources: Digital Green. (2024). Farmer.Chat: Generative AI for small-scale farmers. Digital Green.; digitalgreen.org.Sources use different measurement bases; read the comparison directionally, not as one exact scale.Unitreported users1MLater update830K2024 figure
FarmerChat moved from roughly 830,000 reported users to more than 1 million across five countries. That is meaningful evidence of demand for localized agricultural advice. It is not yet evidence that farms earned more because of generative AI. Usage can scale much faster than causal proof.
FarmerChat superó 1 millón de usuarios reportadosantes de conocerse su impacto en utilidadesAlcance de FarmerChat reportado por la organización en una cifra de 2024 y una actualización posterior. Elvalor más reciente es un mínimo.Made for Freddy Vega, by HanademiFuentes: Digital Green. (2024). Farmer.Chat: Generative AI for small-scale farmers. Digital Green.; digitalgreen.org.Las fuentes usan bases de medición distintas; lea la comparación como tendencia, no como una escala exacta.Unidadusuarios reportados1 MActualización posterior830KCifra de 2024
FarmerChat pasó de cerca de 830.000 usuarios reportados a más de 1 millón en cinco países. Es evidencia importante de demanda por asesoría agrícola localizada. Todavía no prueba que las fincas ganaran más gracias a la IA generativa. El uso puede crecer mucho más rápido que la evidencia causal.
The phone is already in the fieldA woman stands in a crop field holding a smartphone.Sources: Testing AI advisory services – insights from FarmerChat in India and Kenya. rural21.com.
Smartphone delivery can reach crop fields, but its effect on profit still requires evaluation.
El teléfono ya está en el campoUna mujer está de pie en un campo de cultivo sosteniendo un teléfono inteligente.Fuentes: Testing AI advisory services – insights from FarmerChat in India and Kenya. rural21.com.
La entrega por teléfono inteligente puede llegar a los cultivos, pero su efecto en las ganancias aún debe evaluarse.
FarmerChat scaled before its randomizedimpact test beganDated FarmerChat evidence milestones: Kenyan interactions collected in 2023-2024 and randomized-trialregistration published in October 2025.Made for Freddy Vega, by HanademiSources: doi.org.; huggingface.co.; 60decibels.com.Unitcalendar periodOct 1, 2023Apr 30, 2024Oct 26, 2025Oct 1, 2023–Apr 30, 20242,300 Kenyan interactionrecordsOct 26, 2025Randomized-trial registration
A public dataset captured 2,300 anonymized Kenyan interactions collected from October 2023 to April 2024. A randomized trial of FarmerChat's impact on Kenyan smallholders was registered in October 2025. A separate survey of 450 users found trust and many first-time digital-advice users. The next evidence milestone must be income, not another reach count.
FarmerChat escaló antes de iniciar suprueba aleatorizada de impactoHitos fechados de FarmerChat: interacciones kenianas recopiladas en 2023-2024 y registro del ensayoaleatorizado publicado en octubre de 2025.Made for Freddy Vega, by HanademiFuentes: doi.org.; huggingface.co.; 60decibels.com.Unidadperiodo calendarioOct 1, 2023Apr 30, 2024Oct 26, 2025Oct 1, 2023–Apr 30, 20242.300 interacciones de KeniaOct 26, 2025Registro del ensayoaleatorizado
Un conjunto público reunió 2.300 interacciones anonimizadas de Kenia recopiladas entre octubre de 2023 y abril de 2024. Un ensayo aleatorizado sobre el impacto de FarmerChat en pequeños productores kenianos se registró en octubre de 2025. Otra encuesta a 450 usuarios encontró confianza y muchos usuarios nuevos en asesoría digital. El siguiente hito debe ser ingreso, no otro conteo de alcance.
Sub-Saharan internet use reached only35.7% in 2025Individuals using the internet as a share of Sub-Saharan Africa's population, 2005-2025, World Bankseries.Made for Freddy Vega, by HanademiSources: World Bank. (2024). Individuals using the Internet (% of population). World Development Indicators.; World Bank. (n.d.).Individuals using the Internet (% of population) [IT.NET.USER.ZS]. World Bank Open Data.Unit% of population0%10%20%30%2005200820112014201720202023Internet use was 2% in 2005.It reaches 35.7% in 2025.
Internet use in Sub-Saharan Africa rose from 2% in 2005 to 35.7% in 2025. The growth is substantial, but nearly two-thirds of the population still falls outside the measure. Digital-only entrepreneur support will therefore select better-connected users first. Assisted and offline channels remain essential.
El uso de internet subsahariano llegó solo a35,7% en 2025Personas que usan internet como proporción de la población de África subsahariana, 2005-2025, serie delBanco Mundial.Made for Freddy Vega, by HanademiFuentes: World Bank. (2024). Individuals using the Internet (% of population). World Development Indicators.; World Bank. (n.d.).Individuals using the Internet (% of population) [IT.NET.USER.ZS]. World Bank Open Data.Unidad% de la población0 %10 %20 %30 %2005200820112014201720202023El uso de internet era 2% en 2005.Llega a 35,7% en 2025.
El uso de internet en África subsahariana pasó de 2% en 2005 a 35,7% en 2025. El crecimiento es importante, pero casi dos tercios de la población aún quedan fuera de la medida. El apoyo empresarial solo digital seleccionará primero a usuarios mejor conectados. Los canales asistidos y sin conexión siguen siendo esenciales.
Digital access has a storefrontA woman walks past a cyber cafe advertising printing, scanning and photocopying.Sources: Gen AI field experiment shows mixed results in helping small businesses .... newsroom.haas.berkeley.edu.
A storefront access point shows why digital delivery can depend on local intermediaries as well as personal connectivity.
El acceso digital tiene un localUna mujer pasa frente a un cibercafé que anuncia servicios de impresión, escaneo y fotocopiado.Fuentes: Gen AI field experiment shows mixed results in helping small businesses .... newsroom.haas.berkeley.edu.
Un punto de acceso comercial muestra que la entrega digital puede depender tanto de intermediarios locales como de la conectividad personal.
South Asian internet use accelerated after2018Individuals using the internet as a share of South Asia's population, 2005-2025, World Bank series.Made for Freddy Vega, by HanademiSources: World Bank. (n.d.). Individuals using the Internet (% of population) [IT.NET.USER.ZS]. World Bank Open Data.Unit% of population0%20%40%60%2005200820112014201720202023Internet use was 20.9% in2018.It reaches 67.8% in 2025.
South Asia followed a very different adoption curve. Internet use rose from 20.9% in 2018 to 67.8% in 2025. That rapid shift expands the market for digital entrepreneur services in only a few years. Program architecture must therefore adapt to each region's actual access curve.
El uso de internet en Asia meridional seaceleró después de 2018Personas que usan internet como proporción de la población de Asia meridional, 2005-2025, serie delBanco Mundial.Made for Freddy Vega, by HanademiFuentes: World Bank. (n.d.). Individuals using the Internet (% of population) [IT.NET.USER.ZS]. World Bank Open Data.Unidad% de la población0 %20 %40 %60 %2005200820112014201720202023El uso de internet era 20,9% en2018.Llega a 67,8% en 2025.
Asia meridional siguió una curva de adopción muy diferente. El uso de internet pasó de 20,9% en 2018 a 67,8% en 2025. Ese cambio rápido amplía el mercado de servicios empresariales digitales en pocos años. La arquitectura del programa debe adaptarse a la curva real de acceso de cada región.
Latin America narrowed its internet gapwith EuropeIndividuals using the internet as a share of population in Latin America and the Caribbean and Europe andCentral Asia, 2005-2025.Made for Freddy Vega, by HanademiSources: World Bank. (n.d.). Individuals using the Internet (% of population) [IT.NET.USER.ZS]. World Bank Open Data.; GSMA. (2024).The mobile gender gap report 2024. GSMA.Unit% of population0%25%50%75%2005200820112014201720202023Latin Americaand CaribbeanEurope andCentral AsiaLatin America reaches 83.7% in2025.Europe and Central Asia reaches92.4%.
Latin America and the Caribbean moved from 16.6% internet use in 2005 to 83.7% in 2025, narrowing the gap with Europe and Central Asia. Yet access is not evenly distributed inside regional averages. GSMA estimated that women in low and middle-income countries were 15% less likely than men to use mobile internet in 2023. Inclusive design must look below the headline rate.
América Latina redujo su brecha de internetcon EuropaPersonas que usan internet como proporción de la población en América Latina y el Caribe y Europa y Asiacentral, 2005-2025.Made for Freddy Vega, by HanademiFuentes: World Bank. (n.d.). Individuals using the Internet (% of population) [IT.NET.USER.ZS]. World Bank Open Data.; GSMA.(2024). The mobile gender gap report 2024. GSMA.Unidad% de la población0 %25 %50 %75 %2005200820112014201720202023América Latinay el CaribeEuropa y AsiacentralAmérica Latina llega a 83,7% en2025.Europa y Asia central llega a92,4%.
América Latina y el Caribe pasó de 16,6% de uso de internet en 2005 a 83,7% en 2025, reduciendo la brecha con Europa y Asia central. Sin embargo, el acceso no se distribuye de manera uniforme dentro de los promedios regionales. GSMA estimó que las mujeres de países de ingresos bajos y medios tenían 15% menos probabilidad que los hombres de usar internet móvil en 2023. El diseño inclusivo debe mirar debajo de la tasa general.
Sub-Saharan electricity access reachedonly 55% in 2024Population with access to electricity in Sub-Saharan Africa, 2000-2024, World Bank series.Made for Freddy Vega, by HanademiSources: World Bank. (2024). Access to electricity (% of population). World Development Indicators.; World Bank. (n.d.). Access toelectricity (% of population) [EG.ELC.ACCS.ZS]. World Bank Open Data.Unit% of population0%20%40%2000200420082012201620202024Access was 25.7% in 2000.Access reaches 55.1% in 2024.
Electricity access in Sub-Saharan Africa more than doubled from 25.7% in 2000 to 55.1% in 2024. The progress is real, but nearly half the population remains outside the measure. Charging devices and maintaining connectivity depend on this layer. Digital inclusion begins below the software stack.
El acceso eléctrico subsahariano llegó soloa 55% en 2024Población con acceso a electricidad en África subsahariana, 2000-2024, serie del Banco Mundial.Made for Freddy Vega, by HanademiFuentes: World Bank. (2024). Access to electricity (% of population). World Development Indicators.; World Bank. (n.d.). Access toelectricity (% of population) [EG.ELC.ACCS.ZS]. World Bank Open Data.Unidad% de la población0 %20 %40 %2000200420082012201620202024El acceso era 25,7% en 2000.El acceso llega a 55,1% en 2024.
El acceso a electricidad en África subsahariana aumentó de 25,7% en 2000 a 55,1% en 2024. El progreso es real, pero casi la mitad de la población aún queda fuera de la medida. Cargar dispositivos y mantener conexión dependen de esta capa. La inclusión digital comienza debajo del software.
Global electricity access advanced slowlyafter reaching 90%World population with access to electricity, 2000-2024, World Bank series.Made for Freddy Vega, by HanademiSources: World Bank. (n.d.). Access to electricity (% of population) [EG.ELC.ACCS.ZS]. World Bank Open Data.Unit% of population80%90%2000200420082012201620202024Global access crossed 90% in2019.It reaches 91.9% in 2024.
Global electricity access climbed from 78.2% in 2000 to 91.9% in 2024. Progress slowed after the world crossed 90% in 2019. That global average can make digital delivery appear nearly universal. Sub-Saharan Africa's 55.1% rate shows why regional design cannot rely on the average.
El acceso eléctrico mundial avanzólentamente después de llegar a 90%Población mundial con acceso a electricidad, 2000-2024, serie del Banco Mundial.Made for Freddy Vega, by HanademiFuentes: World Bank. (n.d.). Access to electricity (% of population) [EG.ELC.ACCS.ZS]. World Bank Open Data.Unidad% de la población80 %90 %2000200420082012201620202024El acceso mundial superó 90%en 2019.Llega a 91,9% en 2024.
El acceso eléctrico mundial pasó de 78,2% en 2000 a 91,9% en 2024. El progreso se desaceleró después de que el mundo superó 90% en 2019. Ese promedio puede hacer parecer casi universal la distribución digital. La tasa de 55,1% de África subsahariana muestra por qué el diseño regional no puede depender del promedio.
South Asia closed the electricity gap thatLatin America had already solvedPopulation with access to electricity in South Asia and Latin America and the Caribbean, 2000-2024.Made for Freddy Vega, by HanademiSources: World Bank. (n.d.). Access to electricity (% of population) [EG.ELC.ACCS.ZS]. World Bank Open Data.Unit% of population60%80%100%2000200420082012201620202024South AsiaLatin Americaand Caribbean99.8% in 202498.0% in 2024
South Asia began 2000 with electricity access near 57%, far behind Latin America and the Caribbean. By 2024, South Asia reached 99.8% while Latin America and the Caribbean stood at 98.0%. Infrastructure is not a permanent regional trait. Entrepreneur programs should update channel assumptions instead of freezing them at launch.
Asia meridional cerró la brecha eléctricaque América Latina ya había resueltoPoblación con acceso a electricidad en Asia meridional y América Latina y el Caribe, 2000-2024.Made for Freddy Vega, by HanademiFuentes: World Bank. (n.d.). Access to electricity (% of population) [EG.ELC.ACCS.ZS]. World Bank Open Data.Unidad% de la población60 %80 %100 %2000200420082012201620202024AsiameridionalAmérica Latinay el Caribe99,8% en 202498,0% en 2024
Asia meridional comenzó 2000 con acceso eléctrico cercano a 57%, muy por detrás de América Latina y el Caribe. Para 2024, Asia meridional llegó a 99,8%, mientras América Latina y el Caribe alcanzó 98,0%. La infraestructura no es una característica regional permanente. Los programas empresariales deben actualizar sus supuestos de distribución.
Basic-phone AI can use channels above 100subscriptions per 100 peopleMobile cellular subscriptions per 100 people in Ghana, Kenya and Zambia, 2000-2024. Subscriptions mayexceed unique users. Log scale: every gridline is ×10.Made for Freddy Vega, by HanademiSources: Microsoft. (2024). Viamo and Microsoft bring generative AI to people using basic mobile phones. MicrosoftSource.; World Bank. (n.d.). Mobile cellular subscriptions (per 100 people) [IT.CEL.SETS.P2]. World Bank Open Data.Unitsubscriptions per 100 people0.11101001,0002000200420082012201620202024GhanaKenyaZambiaZambia reaches 108.7subscriptions per 100 people.Kenya reaches 126.5subscriptions per 100.
Microsoft and Viamo introduced 321 AI through basic-phone and local-language channels, beginning with Zambia. The distribution layer is substantial: Zambia exceeded 108 subscriptions per 100 people in 2024, while Kenya exceeded 126. These are subscriptions, not unique users. Still, the channel reaches beyond smartphone-only products.
La IA para teléfonos básicos puede usar canales conmás de 100 suscripciones por cada 100 personasSuscripciones móviles por cada 100 personas en Ghana, Kenia y Zambia, 2000-2024. Las suscripcionespueden superar a los usuarios únicos. Escala logarítmica: cada línea es ×10.Made for Freddy Vega, by HanademiFuentes: Microsoft. (2024). Viamo and Microsoft bring generative AI to people using basic mobile phones.Microsoft Source.; World Bank. (n.d.). Mobile cellular subscriptions (per 100 people) [IT.CEL.SETS.P2]. WorldBank Open Data.Unidadsuscripciones por cada 100 personas0,11101001.0002000200420082012201620202024GhanaKeniaZambiaZambia llega a 108,7suscripciones por cada 100personas.Kenia llega a 126,5 suscripcionespor cada 100.
Microsoft y Viamo presentaron 321 AI mediante teléfonos básicos y canales en idiomas locales, comenzando con Zambia. La capa de distribución es amplia: Zambia superó 108 suscripciones por cada 100 personas en 2024 y Kenia superó 126. Son suscripciones, no usuarios únicos. Aun así, el canal llega más allá de productos exclusivos para teléfonos inteligentes.
Design must start with constraintsA man sits in an unfinished brick room beside bundles of plastic bottles and furniture.Sources: No smartphone or internet? No problem; AI-backed phone has the answers | Reuters. reuters.com.
Basic-phone options can extend support into working environments that digital-only programs may overlook.
El diseño debe partir de las limitacionesUn hombre está sentado en una habitación de ladrillo sin terminar junto a fardos de botellas deplástico y muebles.Fuentes: No smartphone or internet? No problem; AI-backed phone has the answers | Reuters. reuters.com.
Las opciones para teléfonos básicos pueden llevar apoyo a entornos de trabajo que los programas exclusivamente digitales podrían pasar por alto.
Mobile reach can reverse sharply evenafter years of growthMobile cellular subscriptions per 100 people in Nigeria and Togo, 2000-2024, World Bank series. Logscale: every gridline is ×10.Made for Freddy Vega, by HanademiSources: World Bank. (n.d.). Mobile cellular subscriptions (per 100 people) [IT.CEL.SETS.P2]. World Bank Open Data.Unitsubscriptions per 100 people0.010.11101002000200420082012201620202024NigeriaTogoNigeria reached 99.6subscriptions per 100 in 2022.Nigeria fell to 70.8 in 2024.Togo reached 80.8 in 2024.
Nigeria and Togo both built broad mobile channels from almost nothing. Nigeria reached nearly 100 subscriptions per 100 people in 2022, then fell to 70.8 in 2024. Togo reached 80.8. A distribution plan cannot treat one penetration figure as permanent infrastructure.
El alcance móvil puede retroceder confuerza tras años de crecimientoSuscripciones móviles por cada 100 personas en Nigeria y Togo, 2000-2024, series del Banco Mundial.Escala logarítmica: cada línea es ×10.Made for Freddy Vega, by HanademiFuentes: World Bank. (n.d.). Mobile cellular subscriptions (per 100 people) [IT.CEL.SETS.P2]. World Bank OpenData.Unidadsuscripciones por cada 100 personas0,010,11101002000200420082012201620202024NigeriaTogoNigeria llegó a 99,6 suscripcionespor cada 100 en 2022.Nigeria cayó a 70,8 en 2024.Togo llegó a 80,8 en 2024.
Nigeria y Togo construyeron amplios canales móviles desde niveles casi nulos. Nigeria llegó a casi 100 suscripciones por cada 100 personas en 2022 y luego cayó a 70,8 en 2024. Togo llegó a 80,8. Un plan de distribución no puede tratar una cifra de penetración como infraestructura permanente.
Malawi's mobile channel expanded, butUlangizi remains a pilotMobile subscriptions per 100 people in Malawi, 2000-2024, with Opportunity International's Ulangiziagricultural assistant as the relevant pilot. Log scale: every gridline is ×10.Made for Freddy Vega, by HanademiSources: Opportunity International. (2024). Ulangizi: An AI-powered farming assistant for Malawi. OpportunityInternational.; World Bank. (n.d.). Mobile cellular subscriptions (per 100 people) [IT.CEL.SETS.P2]. World Bank OpenData.Unitsubscriptions per 100 people0.11101002000200420082012201620202024Malawi had 0.4 subscriptions per100 people.Malawi reaches 69.3subscriptions per 100.
Opportunity International describes Ulangizi as an AI agricultural assistant piloted with smallholders in Malawi. The country's mobile channel rose from 0.4 subscriptions per 100 people in 2000 to 69.3 in 2024. That makes assisted digital delivery increasingly plausible. The pilot still needs evidence on farm profit and harm.
El canal móvil de Malawi se expandió, peroUlangizi sigue siendo un pilotoSuscripciones móviles por cada 100 personas en Malawi, 2000-2024, con el asistente agrícola Ulangizi deOpportunity International como piloto relevante. Escala logarítmica: cada línea es ×10.Made for Freddy Vega, by HanademiFuentes: Opportunity International. (2024). Ulangizi: An AI-powered farming assistant for Malawi. OpportunityInternational.; World Bank. (n.d.). Mobile cellular subscriptions (per 100 people) [IT.CEL.SETS.P2]. World BankOpen Data.Unidadsuscripciones por cada 100 personas0,11101002000200420082012201620202024Malawi tenía 0,4 suscripcionespor cada 100 personas.Malawi llega a 69,3 suscripcionespor cada 100.
Opportunity International describe Ulangizi como un asistente agrícola de IA probado con pequeños productores en Malawi. El canal móvil del país pasó de 0,4 suscripciones por cada 100 personas en 2000 a 69,3 en 2024. Eso hace cada vez más viable la distribución digital asistida. El piloto aún necesita evidencia sobre utilidades agrícolas y daños.
Farmerline reports 2 million farmers in aneconomy still tied to agricultureAgricultural employment share in Côte d'Ivoire, modeled ILO estimate, 1991-2025. Farmerline's reach iscompany-reported and may mix registered and active users.Made for Freddy Vega, by HanademiSources: Farmerline. (2024). Company impact and reach. Farmerline.; World Bank. (n.d.). Employment in agriculture (% of totalemployment) (modeled ILO estimate) [SL.AGR.EMPL.ZS]. World Bank Open Data.Unit% of employment45%50%1991199620012006201120162021Agricultural employmentpeaked at 52.9% in 2016.It remains 45.4% in 2025.
Farmerline reports reaching more than 2 million farmers through locally built agricultural distribution. Côte d'Ivoire's modeled agricultural employment share remains 45.4% in 2025, showing the scale of the relevant livelihood base. Public reach figures do not consistently separate registered from active users. Distribution becomes valuable when participation and outcomes can be verified.
Farmerline reporta 2 millones de agricultores en unaeconomía aún ligada a la agriculturaProporción del empleo agrícola en Côte d'Ivoire, estimación modelada de la OIT, 1991-2025. El alcancede Farmerline es reportado por la empresa y puede mezclar usuarios registrados y activos.Made for Freddy Vega, by HanademiFuentes: Farmerline. (2024). Company impact and reach. Farmerline.; World Bank. (n.d.). Employment in agriculture (% of totalemployment) (modeled ILO estimate) [SL.AGR.EMPL.ZS]. World Bank Open Data.Unidad% del empleo45 %50 %1991199620012006201120162021El empleo agrícola alcanzó unpico de 52,9% en 2016.Permanece en 45,4% en 2025.
Farmerline reporta haber alcanzado a más de 2 millones de agricultores mediante distribución agrícola construida localmente. La proporción modelada de empleo agrícola en Côte d'Ivoire permanece en 45,4% en 2025, lo que muestra el tamaño de la base relevante. Las cifras públicas no separan consistentemente usuarios registrados de activos. La distribución se vuelve valiosa cuando pueden verificarse participación y resultados.
Amini raised $6 million around an economystill exposed to agricultureAgriculture, forestry and fishing as a share of Cameroon's GDP, 2000-2025, beside Amini's announced $2million pre-seed and $4 million seed rounds.Made for Freddy Vega, by HanademiSources: Amini. (2024). Kate Kallot, founder and chief executive officer. Amini.; TechCrunch. (2023). Amini raises $4Mseed funding for African climate-data infrastructure. TechCrunch.; World Bank. (n.d.). Agriculture, forestry, and fishing,value added (% of GDP) [NV.AGR.TOTL.ZS]. World Bank Open Data.UnitGDP share and USD millionsCameroon agricultural exposure18.1%16.8%20002025Amini financing rounds$2M$4MPre-seedSeed, 2023
Cameroon-born Kate Kallot founded Amini after leadership roles linked to NVIDIA, Arm and Intel. Amini announced $2 million in pre-seed funding and a $4 million seed round in 2023. Cameroon still derives a meaningful share of GDP from agriculture. Better environmental data may help credit and insurance, but funding the mapmaker does not prove farmers became richer.
Amini recaudó 6 millones de dólares en unaeconomía aún expuesta a la agriculturaAgricultura, silvicultura y pesca como proporción del PIB de Camerún, 2000-2025, junto con las rondasanunciadas de Amini por 2 millones y 4 millones de dólares.Made for Freddy Vega, by HanademiFuentes: Amini. (2024). Kate Kallot, founder and chief executive officer. Amini.; TechCrunch. (2023). Amini raises$4M seed funding for African climate-data infrastructure. TechCrunch.; World Bank. (n.d.). Agriculture, forestry,and fishing, value added (% of GDP) [NV.AGR.TOTL.ZS]. World Bank Open Data.Unidadproporción del PIB y millones de USDExposición agrícola de Camerún18,1 %16,8 %20002025Rondas de financiamiento de Amini$2 M$4 MPresemillaSemilla, 2023
Kate Kallot, nacida en Camerún, fundó Amini tras cargos de liderazgo vinculados con NVIDIA, Arm e Intel. Amini anunció 2 millones de dólares de financiamiento presemilla y una ronda semilla de 4 millones en 2023. Camerún todavía obtiene una parte importante del PIB de la agricultura. Mejores datos ambientales pueden apoyar crédito y seguros, pero financiar a quien crea los mapas no prueba que los agricultores se enriquecieron.
Data must reach the fieldA farmer examines corn plants while holding a tablet in a field.Sources: 60decibels.com.
Agricultural data infrastructure creates value only when it improves decisions made in real operating environments.
Los datos deben llegar al campoUn agricultor examina plantas de maíz mientras sostiene una tableta en un campo.Fuentes: 60decibels.com.
La infraestructura de datos agrícolas solo crea valor cuando mejora las decisiones tomadas en entornos reales de trabajo.
One quarter of Ghana's young womenremained outside work and educationFemale youth not in education, employment or training as a share of Ghana's female youth population,available observations from 2000-2024.Made for Freddy Vega, by HanademiSources: World Bank. (n.d.). Share of youth not in education, employment or training, female (% of female youth population)[SL.UEM.NEET.FE.ZS]. World Bank Open Data.Unit% of female youth0%10%20%30%20002010201520212023The series reached a low of 17.1%in 2013.It peaked at 37.1% in 2015.The 2024 value is 26.3%.33.2%
Ghana's female youth exclusion rate has moved sharply over time. It fell to 17.1% in 2013, spiked to 37.1% in 2015 and stood at 26.3% in 2024. That is more than one in four young women outside education, employment or training. Entrepreneur support can matter most before formal labor-market attachment exists.
Una cuarta parte de las jóvenes de Ghanaseguía fuera del trabajo y la educaciónMujeres jóvenes sin educación, empleo ni capacitación como proporción de la población femenina joven deGhana, observaciones disponibles de 2000-2024.Made for Freddy Vega, by HanademiFuentes: World Bank. (n.d.). Share of youth not in education, employment or training, female (% of female youth population)[SL.UEM.NEET.FE.ZS]. World Bank Open Data.Unidad% de mujeres jóvenes0 %10 %20 %30 %20002010201520212023La serie alcanzó un mínimo de17,1% en 2013.Alcanzó un pico de 37,1% en 2015.El valor de 2024 es 26,3%.33,2 %
La tasa de exclusión de mujeres jóvenes en Ghana ha cambiado con fuerza. Cayó a 17,1% en 2013, subió a 37,1% en 2015 y se ubicó en 26,3% en 2024. Es más de una de cada cuatro mujeres jóvenes fuera de educación, empleo o capacitación. El apoyo empresarial puede ser más importante antes de existir un vínculo con el mercado laboral formal.
Fund human capability alongside AIEstimated profit effects in a randomized Togo study: personal-initiative training versus conventionalbusiness training.Made for Freddy Vega, by HanademiSources: Campos, F., et al. (2017). Teaching personal initiative beats traditional training in boosting small business in WestAfrica. Science.Unitestimated profit effectPersonal initiative30%Conventional training11%2.7x
A randomized Togo study estimated that personal-initiative training raised profits about 30%. Conventional business training produced an estimated 11% gain. The difference makes entrepreneur capability a serious intervention target. A defensible AI program should compare AI alone, AI with human coaching and a control group for at least one year.
Financie capacidades humanas junto con IAEfectos estimados sobre utilidades en un estudio aleatorizado en Togo: capacitación en iniciativa personalfrente a capacitación empresarial convencional.Made for Freddy Vega, by HanademiFuentes: Campos, F., et al. (2017). Teaching personal initiative beats traditional training in boosting small businessin West Africa. Science.Unidadefecto estimado sobre utilidadesIniciativa personal30 %Capacitación convencional11 %2,7x
Un estudio aleatorizado en Togo estimó que la capacitación en iniciativa personal elevó las utilidades cerca de 30%. La capacitación empresarial convencional produjo un aumento estimado de 11%. La diferencia convierte las capacidades del emprendedor en un objetivo serio de intervención. Un programa defendible debe comparar IA sola, IA con acompañamiento humano y un grupo de control durante al menos un año.
AI support needs capability safeguards,outcome measurement and human guidance.The evidence is mixed: one study estimated an 8% decline for weaker entrepreneurs,another found a 19-point accuracy penalty beyond the reliable frontier, 7,000workplaces showed no significant earnings or hours gain, and personal-initiative trainingraised profits about 30% versus 11% for conventional training.Sources: Otis, N., Clarke, R., Delecourt, S., Holtz, D., & Koning, R. (2024). The uneven impact of generative AI on entrepreneurial performance. Harvard Business School Working Paper.; TheUneven Impact of Generative AI on Entrepreneurial ....; Dell'Acqua, F., et al. (2023). Navigating the jagged technological frontier. Harvard Business School Working Paper 24-013.; What isthe jagged AI frontier?.; fortune.com.; Campos, F., et al. (2017). Teaching personal initiative beats traditional training in boosting small business in West Africa. Science.
The market is vast, informal and unevenly connected. AI can improve bounded work, yet the same tool can hurt less-prepared users or fail outside its reliable frontier. Programs already show impressive reach, while causal income evidence remains scarce. The strongest proposal combines measured outcomes, human support and safeguards.
El apoyo con IA necesita salvaguardias decapacidad, medición de resultados yorientación humana.La evidencia es mixta: un estudio estimó una caída del 8% para emprendedores másdébiles, otro encontró una penalización de 19 puntos en precisión fuera de la fronteraconfiable, 7.000 lugares de trabajo no mostraron aumentos significativos de ingresos nihoras, y la capacitación en iniciativa personal elevó las utilidades cerca de 30% frente a11% con capacitación convencional.Fuentes: Otis, N., Clarke, R., Delecourt, S., Holtz, D., & Koning, R. (2024). The uneven impact of generative AI on entrepreneurial performance. Harvard Business School Working Paper.; TheUneven Impact of Generative AI on Entrepreneurial ....; Dell'Acqua, F., et al. (2023). Navigating the jagged technological frontier. Harvard Business School Working Paper 24-013.; What isthe jagged AI frontier?.; fortune.com.; Campos, F., et al. (2017). Teaching personal initiative beats traditional training in boosting small business in West Africa. Science.
El mercado es enorme, informal y tiene conectividad desigual. La IA puede mejorar tareas delimitadas, pero la misma herramienta puede perjudicar a usuarios menos preparados o fallar fuera de su frontera confiable. Los programas ya muestran un alcance impresionante, mientras la evidencia causal sobre ingresos sigue siendo escasa. La propuesta más sólida combina resultados medidos, apoyo humano y salvaguardias.
In summaryMade for Freddy Vega, by HanademiSources: Otis, N., Clarke, R., Delecourt, S., Holtz, D., & Koning, R. (2024). The uneven impact of generative AI on entrepreneurial performance. HarvardBusiness School Working Paper.; doi.org.; TechCrunch. (2023). Amini raises $4M seed funding for African climate-data infrastructure. TechCrunch.Lower baseline performers offered the same AI mentor experienced an estimated performance decline of approximately 8%.An AEA randomized-trial registration for Farmer.Chat’s impact on Kenyan smallholders was published on October 26, 2025.Amini announced a $4 million seed round in 2023 after an earlier $2 million pre-seed round.
The value of the research is not only what each source knew, but what became visible when their evidence was combined.
En resumenMade for Freddy Vega, by HanademiFuentes: Otis, N., Clarke, R., Delecourt, S., Holtz, D., & Koning, R. (2024). The uneven impact of generative AI on entrepreneurial performance. HarvardBusiness School Working Paper.; doi.org.; TechCrunch. (2023). Amini raises $4M seed funding for African climate-data infrastructure. TechCrunch.Los participantes con menor desempeño inicial que recibieron el mismo mentor de IA experimentaron una caída estimadade aproximadamente 8%.El 26 de octubre de 2025 se publicó en la AEA el registro de un ensayo aleatorizado sobre el impacto de Farmer.Chat enpequeños productores de Kenia.Amini anunció una ronda semilla de 4 millones de dólares en 2023 después de una ronda presemilla anterior de 2 millones.
El valor de la investigación no está solo en cada fuente, sino en lo que apareció al combinar sus evidencias.

The research behind this deck

User capability can reverse the effect of the same AI product. 3 technical terms carry much of the argument.

Key findings

The argument

This research is published in English and Spanish. Ver en español

La investigación detrás de esta presentación

La capacidad del usuario puede invertir el efecto del mismo producto de IA. 3 términos técnicos sostienen buena parte del argumento.

Hallazgos clave

El argumento

Esta investigación se publica en inglés y español. Read in English