In June 2011, I wrote that microtasking gave youth a value chain they could use to climb out of poverty. On April 16, 2026, Sama issued redundancy notices to 1,108 Nairobi staff because one client, Meta, ended one engagement.

Fifteen years separate those two sentences. In between, the Rockefeller Foundation and the International Finance Corporation treated impact sourcing as a development strategy. Neither priced the risk that ended it in Nairobi.

Sign Up Now for more digital jobs insights

Why Development Organizations Bought In

The pitch was simple, and I bought it too. Hire young people from poor neighborhoods to do digital work for global clients, pay them formal wages, train them, and watch them move up.

In May 2013, Rockefeller Foundation president Judith Rodin pledged nearly $100 million to Digital Jobs Africa. The stated goal was one million lives in six countries, partly by growing the impact sourcing sector. At a Technology Salon that July, Rockefeller’s Lillian Chege said some programs had placed 100% of participating youth in formal jobs.

Rockefeller also funded Samasource directly. A World Bank-hosted Solutions for Youth Employment case study lists Rockefeller grants to Samasource in Kenya totaling about $1.6 million across three phases between 2009 and 2015.

In March 2017, Samasource told a New York Technology Salon that its workers saw a fourfold income increase after four years, and that it wanted workers to move on within two years. The same Salon recorded two warnings.

• Samasource said donors had heavily subsidized it for eight years.
• Clients told Banyan Global’s researchers: impact was “the cherry on top.”

The theory was a ladder. Workers would build skills at Sama and leave for better jobs. Their seats would go to the next cohort.

Evidence was Workers Leaving Individually

MIT economists David Atkin and Antoinette Schoar ran the test the sector needed. With Innovations for Poverty Action, they randomized 1,115 Nairobi youth who applied to Sama’s AI 101 training.

Youth who got training plus a job referral to Sama’s delivery center earned 37% more per month than the comparison group, and their unemployment rate was 10 percentage points lower. Women earned 60% more. Youth who got the same training without the referral showed no gain in earnings or employment after 2.5 years.

The full paper puts delivery-center pay at about $301 a month with benefits, roughly 2.5 times Kenya’s formal minimum wage, and average agent tenure at 25 months in 2020.

Participants who worked at the delivery center and then left still reported higher earnings at endline, and the authors read that as experience with value elsewhere. That comparison is not randomized, because the people who leave Sama are not a random draw. It is still the best evidence that the ladder worked for some climbers.

Those exits happened one worker at a time, in a study period when a large client contract had Sama hiring. Nobody measured what happens when 1,108 workers with the same skills hit the Nairobi job market in the same month, after the buyer of their work took it elsewhere.

Sama’s own alumni data cannot fill that gap. ImpactMatters audited Samasource’s evidence in 2016 and found that only 32% of sampled alumni completed the company’s follow-up survey. The Solutions for Youth Employment case study rests its claim that 85% of former workers kept working or studying on that same survey design.

One Client Shaped Sama for Nine Years

Meta first contracted Sama in 2017 for data labelling, with about 1,500 staff on the work. When Daniel Motaung led an attempt to unionize Sama’s Facebook moderators in 2019, TIME reported his dismissal letter said his actions put Sama’s relationship with Facebook at great risk. Sama says it fired Motaung for bullying and coercing colleagues.

In early 2022, Sama dropped its OpenAI work. TIME reported that most of the roughly three dozen workers on it moved to lower-paying workstreams, and others lost their jobs. In January 2023, Sama exited content moderation and cut about 200 jobs, roughly 3% of staff, while keeping the Meta labelling work.

In April 2026, Meta ended that work too. Meta told the BBC it ended the relationship because Sama did not meet its standards. Sama rejects that.

Naftali Wambalo of the Africa Tech Workers Movement told the BBC he believes Meta acted after workers told Swedish reporters they had reviewed intimate footage from Meta’s smart glasses. Mercy Mutemi, the lawyer for former moderators suing Meta and Sama, said Kenya should hear Meta’s statement as a warning: “This is a very flimsy foundation to build your entire industry on.”

Sama’s Kenya country lead, Annepeace Alwala, told TechCabal that client programs evolve as a standard feature of the industry. She is right, and the sector sold that feature. Banyan Global’s 2017 report on microwork and impact sourcing, lists helping companies absorb demand fluctuations among impact sourcing’s benefits to clients. The flexibility clients bought was the volatility workers carried.

Sama’s workers knew it.

In 2023, Oxford Internet Institute researchers working with Fairwork conducted 46 interviews at Sama and found many workers on one-month or three-month contracts. One worker told them complaints about fatigue were met with a reminder that a client exit would mean unemployment.

Sama’s fix tied worker contracts to client contracts: a default of 12 months, or the length of the client contract when the client has no termination rights, unless Sama expects that contract to end early. Fairwork scored Sama 5 out of 10 and noted that its ethical reputation helped it win investment and clients.

Mophat Okinyi calls this the new face of worker exploitation in 2025. I think the problem is duller and easier to fix. Nobody with money in the deal measured what happens when a core client exits and mass layoffs ensue.

IFC Was Ready to Call It Development

In December 2022, IFC disclosed a proposed investment of up to $10 million in Sama to expand in Uganda. Up to $5 million was to come from a subordinated co-investment through the IDA20 Private Sector Window Blended Finance Facility, which IFC estimated at an 8.77% subsidy. IFC’s development case promised improved livelihoods and skills upgrading for women and disadvantaged youth.

IFC never put the money in, and it has not said why.

The proposal still matters, because it shows how the sector’s most sophisticated development investor looked at Sama. IFC’s environmental and social review, counted 4,731 Sama staff, about 78% of them in Kenya, and listed content moderation among Sama’s operations.

IFC staff visited Sama sites in Kampala, Gulu, and Nairobi in September 2022, four months after Motaung filed a lawsuit alleging forced labor and inadequate psychosocial support. The review covers fire detection, electrical fire risk, evacuation signage, and workstation equipment.

A text search found no mention of the lawsuit, a court, trauma, psychosocial support, a union, Meta, or client concentration. It describes Sama’s strategy as serving large steady clients.

IFC did see one risk clearly.

Its draft action plan required Sama to update its redundancy procedures with an analysis of alternatives to retrenchment and a plan that engages workers before and during layoffs. The development bank knew the core client exit scenario needed a plan.

Formal Employment is a Benefit

J-PAL’s comparison group faced 30% unemployment, and informal jobs held 83.6% of Kenyan employment in 2018, per the same J-PAL summary.

The 1,108 workers got notices under Section 40 of Kenya’s Employment Act, a protection an informal job would not have given them. And the J-PAL paper shows other former Sama workers doing better than a comparison group.

Every BPO firm also carries client concentration risk. But ordinary BPO firms did not ask donors and development banks for concessional capital on a promise of poverty exit and skills transfer. Impact sourcing did. The subsidy bought a claim that no published evidence tests against a mass layoff.

Sama Should Publish the Exit

In late 2023, Sama told Fairwork it would research the job market for its employees and define pathways beyond the firm. It has not published that research. Its 2024 impact report lists an alumni community and career fairs, and claims a 148% average income lift three years after joining Sama, with no methodology.

Sama now holds the cleanest test impact sourcing will ever get: 1,108 workers who left on the same date from the same city. Track them for a year, and publish where they landed and what they earn.

• If the ladder holds under a mass layoff, Sama will own the strongest evidence the sector has ever produced.
• If it fails, every funder weighing the next impact sourcing deal needs to know.

Sama promised to study when staff exit. It should start with the 1,108.

The post IFC Checked Sama’s Fire Exits. Nobody Checked the Client Exit. appeared first on ICTworks.

Private investors put 6% of healthcare investment into women’s health, according to the World Economic Forum’s Women’s Health Investment Outlook. The report found that 90% of that money lands in just three clinical areas. WEF says low- and middle-income countries take the hardest hit from the gap.

$100,000 FemTech Opportunity

UNICEF Femtech Ventures program now seeks applicants with AI, machine learning, blockchain and data science applications in three pillars:

• Healthcare access for women and girls, from antenatal care through a child’s first 1,000 days
• Sexual and reproductive health, including HPV screening and consent-preserving data tools
• Mental wellbeing and safety, including gender-based violence response and maternal mental health screening inside DHIS2

Each selected company receives up to US$100,000 in equity-free funding and 12 months of technical mentorship. For the first cohort, the Venture Fund lists $640,000 invested across 10 companies, about $64,000 per company.

UNICEF encourages women-led startups and founders under 35 to apply.

Apply Now: Deadline is November 4, 2026

More Funding Opportunities

Please sign up now to get our email updates. Learn how to get startup funding for your technology business, and find new funding opportunities with donors.

We are constantly publishing fundraising guidance and competitive insights like this:

• How I Built an Open Source Digital Marketplace for Women in Costa Rica
• Please Treat Women as Digital Navigators of Facebook Applications
• Gender Productivity Paradox: Artificial Intelligence Is Failing Women

The post Apply Now: $100,000 for Femtech Startups in Emerging Markets appeared first on ICTworks.

On October 18, Anthropic said it would commit 5% of its shares to a new Anthropic Public Benefit Fund, an LLC that could make large donations to nonprofits or direct distributions to specific populations.

Elon Musk replied on X that he recommended “giving the shares directly to the public,” and pledged one of his SpaceX shares to every kid in America next year. That $146 per share gift for 72 million children is around $10 billion. Compared to Musk’s net worth at $904 billion, somewhere a man is very proud of a tip.

So let us take Musk at his word, always a bold financial strategy, and make the pledge global. The problem will be familiar. We have not yet built the capacity to absorb serious philanthropy. Money moves only as fast as the receiving infrastructure. Musk is not the constraint. The pipe is.

Sign Up Now for more digital philanthropy insights

Musk is Right About Direct Gifts

Give the man his due before we check his math. Anthropic’s fund is a wholly owned LLC that its own board directs. A share in a child’s name is an asset the child owns, and nobody in San Francisco decides what it is for.

Cash-transfer researchers have argued for two decades that people living in poverty know what they need better than intermediaries do. Musk has stumbled into their position, which may be the first time he and GiveDirectly agreed on anything.

The trouble is delivery, and delivery is where Musk’s promises typically go to die.

Every Child on Earth is Totally Possible

Roughly 2.3 billion people are under 18. At $146 a share, a global gift costs about $350 billion. That is about 39% of his wealth, and it would leave Musk with a little over $500 billion to get by on.

At the same time, his SpaceX voting power falls from about 82% to about 72% due to its dual shareholder classes. He would still run SpaceX with room to spare.

Narrow it to children born from 2026 through 2028, the same cohorts that get federal Trump Accounts. Call it 400 million children, per the UN, and that’s $58 billion, about 6% of his net worth. Musk could cover that from his Class A shares alone and lose about half a point of voting power. He has lost more value than that in a bad week of posting on X.

Congress Built the Pipe to American Kids

The American gift has a delivery mechanism.

Trump Accounts give every child born from 2025 to 2028 a federally funded $1,000 investment account. In July, SpaceX President Gwynne Shotwell and her husband pledged a share to each of roughly 2 million children already signed up. “Directly to the public” turns out to mean through a federal program Congress has already built. A familiar path for SpaceX.

A newborn in Lilongwe or Kinshasa has no such account and no way to hold a US-listed share. So route the money the boring way that works. Musk gives the stock to the Musk Foundation, the foundation sells it, and it sends cash to GiveDirectly.

Required: Real Musk Foundation Actions

This is the part where the thought experiment requires the most imagination.

The Musk Foundation missed the IRS 5% minimum payout in 2021, 2022 and 2023, and fell $421 million short in 2023. It has no hired staff and a three-member volunteer board that includes Musk.

A $58 billion gift breaks that habit by law.

The 5% rule means about $2.9 billion a year in grants, six times what the foundation gave in 2024. Musk already controls well over 20% of SpaceX’s votes, so the excess business holdings rule should require the foundation to sell the gifted shares within five years.

For once, the foundation would have to sell stock, write checks to strangers, and hire people to read grant reports. Selling about 80 million shares a year is small next to the 555 million shares SpaceX sold at its IPO, so the stock price would survive the experience. The volunteer board might not.

GiveDirectly Would Grow Tenfold

GiveDirectly expects to deliver $90 million in lump-sum transfers in 2025 and says it could responsibly absorb another $250 million or more within a year. Its stretch goal is $5 billion by 2035.

Paying 400 million children before they turn 18 means moving about $3.2 billion a year, roughly ten times GiveDirectly’s stated ceiling. At the ceiling, it takes about 170 years. Musk’s Mars colony might open first.

The footprint is smaller than the cohort, too. GiveDirectly works mostly in sub-Saharan Africa. India, with about 23 million births a year, and China, with 8.7 million, sit outside it. Africa had roughly 47 million births a year. A three-year African cohort is about 140 million children, or $21 billion.

At GiveDirectly’s ceiling, that is still about 60 years of work.

There is one more trap. A single donor supplying most of a charity’s revenue can push it toward failing the IRS public support test. A foundation that writes GiveDirectly a $2.9 billion check to meet its payout could hand it a tax problem along with the money.

Musk Fed the Global Pipe Into the Wood Chipper

The irony writes itself. In February 2025, Musk posted that he had spent the weekend “feeding USAID into the wood chipper.” Yet that very same USAID disbursed $72 billion in fiscal year 2023, much of it through the same receiving systems a global share gift would now need.

The richest man on Earth now wants to give children money, and the pipe he would need is the one he shredded.

We Can Build It Back Better

Here is the good news, and it is real.

GiveDirectly calls its Poverty Relief program shovel-ready and underfunded, with capacity sitting idle this year. Since January 2024 it has run Rx Kids in Flint, Michigan, with the city and the state, sending cash to mothers and infants.

GiveDirectly also reports early evidence that cash during pregnancy and infancy could cut child mortality in half, a claim worth watching as the trial results are published.

A tweeted pledge is the cheap part. The expensive, unglamorous part is the grants office, the payment rail, and the local partner that can take money without breaking. The pipes are there. Is the will?

The post Can Elon Musk Give a SpaceX Share to Every Child on Earth? appeared first on ICTworks.

Mobile money will not grow much further until people use it to pay their utility bills.

That is where digital development’s clearest success story is stuck. Across Africa, people use mobile money every day to send money to family. When the electricity or water bill comes due, many of them still pay in cash.

The industry thinks it knows the fix.

• GSMA Director General Vivek Badrinath names “strengthening consumer protection and fraud controls” as a priority in the State of the Industry Report on Mobile Money 2026.
• The GSMA’s launch announcement adds a call for digital financial literacy programs.
• The Alliance for Financial Inclusion argues that better redress builds consumer confidence.
• We made a version of that case on ICTworks when we argued for accessible grievance redress in digital financial services.

All are wrong. Consumer protection and literacy will not move bill payments. Two things will. People need proof that the payment reached their utility account, and they need to pay less to send it.

Sign Up Now for more digital financial services insights

Utility Bills Are Where Mobile Money Stalls

The GSMA’s 2026 report shows merchant payments jumped 42% to $155 billion in 2025. Bill payments grew 8% to almost $99 billion. In Sub-Saharan Africa, home to over two-thirds of mobile money bill payments, bill value grew just 6%.

Samora Kariuki of Frontier Fintech went through five years of GSMA data. He found bills shrank from 5% of monthly mobile money value in 2020 to 4% in 2025. The World Bank’s Global Findex 2025 found that only 40% of adults in low- and middle-income economies who paid a utility bill paid it digitally.

Six in ten people who paid a utility bill did not use a digital payment.

This matters for growth. Kariuki argues that mobile money mostly moves value between two cash endpoints. A utility bill is one of the few payments where mobile money replaces the cash outright, and every household gets one every month.

Consumer Protection is Not the Gap

A new study from KAIST surveyed 270 mobile money users in 14 African countries. It asked why they skip mobile money for big payments.

Users who trusted refund procedures and customer support felt a little less afraid of losing money. That trust did nothing for their uncertainty about whether the payment went through. Of everything the researchers measured, trust in safeguards had the weakest link to avoiding mobile money.

The authors call this the insurance paradox. A customer can believe the telco will refund a failed payment and still not know whether the electricity company recorded it. It is one small online survey. The direction still matches what the GSMA’s own bill numbers show.

Digital Literacy Is Not the Gap

The KAIST respondents are among the most digitally capable users in Africa. Over nine in ten hold a university degree, and most use mobile money every week.

They still avoid mobile money for big bills.

The biggest driver in the study was a chain. One survey question asked users whether the system forces them through too many steps just to get proof of a bill payment. They agreed that too many steps to finish a payment led to not knowing whether it worked, and that uncertainty led to paying in cash.

Literacy training cannot teach a university graduate to trust a confirmation the utility never sends.

Confidence Has to Come From the Biller

Pay cash at the utility office and you leave with a stamped receipt from the company that took your money. Pay by mobile money and you get an SMS from the telco. Whether the utility credited your account is a separate question, and the customer often cannot answer it.

One bad experience makes this worse. Some KAIST respondents had lived through a failed or wrong transaction. For them, uncertainty pushed them toward cash about three times as hard as it pushed everyone else.

Tanzania’s Ministry of Finance built one answer.

Since a 2017 amendment to the Public Finance Act, all public money flows through the Government electronic Payment Gateway. The agency issues a control number tied to a specific bill. The payer pays it by mobile money or bank, and the payment lands against the agency’s own record. The payee confirms the payment.

I have not found published data on whether GePG moved payments off cash. The design still points at the right institution. Utilities and revenue authorities hold the ledger, so they have to issue the proof.

Price Has to Fall Too

Kariuki’s analysis makes the cost case. Brazil’s free Pix system sees 94% of accounts used monthly and India’s UPI about 70%, against 26% for mobile money. The GSMA itself reports that government transaction taxes push users back to cash. Operators set fees and finance ministries set taxes, so both own this half of the problem.

Price alone will not do it. Kariuki notes that M-Pesa charges fees and still sees near-universal monthly use in Kenya. A cheap payment the utility does not recognize is still a gamble with the household’s electricity.

Fix the receipt and the price together, and the bill moves. Until then, the electricity bill stays a cash transaction and mobile money will underperform its promise.

The post The Path to Mobile Money Growth Goes Through Utility Bills appeared first on ICTworks.

In 2024, 79% of adults worldwide held a financial account. Only 56% could reliably find extra money in an emergency. The World Bank’s Global Findex 2025 reports both numbers, and the 23-point gap between them is the problem MIT Solve wants applicants to close.

The Funding Opportunity

The 2027 Global Economic Prosperity Challenge wants technology-based solutions in these areas:

• Financial health and resilience for individuals, families, and small businesses
• Pathways to quality employment, especially for young people and workers outside formal labor markets as AI changes entry-level jobs
• Safe, affordable, climate-resilient housing for informally settled or displaced people

Solve defines technology broadly. It accepts SMS and AI tools alongside traditional and ancestral building methods.

This is a prize and accelerator program, not a grant. Every selected team receives a $10,000 Solver Award, and Solve reports an average of $40,000 per team after selection. Selected teams also compete for partner prizes, including:

• AI for Humanity Prize: up to $150,000 for AI and data science solutions
• AI for Social Innovation Prize: up to $100,000, split among as many as four teams
• Innovation Prize: $100,000 to one team

Solvers join a nine-month program of coaching and pro bono support, about 25 hours in total. Solve also expects four days in Cambridge, Massachusetts in April and four days in New York City in September.

Apply Now! Deadline is November 2, 2026

More Funding Opportunities

Please sign up now to get our email updates. Learn how to get startup funding for your technology business, and find new funding opportunities with donors.

We are constantly publishing fundraising guidance and competitive insights like this:

• Artificial Intelligence: Meet Indian Loom Artisans
• Mobile Phone Revolutions Make Telecenters Even More Essential
• Shifting the Approach: From Livelihood to Entrepreneurship

The post Apply Now: $150,000 for Economic Prosperity Digital Solutions appeared first on ICTworks.

Imagine this: one founder, typing alone on a Tuesday night, building an agentic swarm to solve a social problem in the United States.

While she sleeps, her agents read 4,000 Medicaid renewal packets, flag the 600 applications missing a pay stub, draft the exemption forms, and queue reminders through a community health center’s patient app.

By morning, a few hundred families who would have lost coverage over a missing form now have the form and can keep their coverage .

Run that every night for a year and she keeps about 115,000 people covered. That is a billion dollars of health care at what Medicaid spends per enrollee, delivered by an organization with one employee.

In May 2025, Anthropic CEO Dario Amodei was asked when the first billion-dollar company with one human employee would appear. He answered “2026.” Silicon Valley has been chasing that company ever since. I think the social sector can build its version first, and I think it should.

The obstacle sits with funders and boards.

Most of them read an organization’s capacity off its staff count. A grant budget built on full-time positions cannot see an organization with one full time staff. The first solo AI nonprofit needs a funder willing to look past the org chart and a motivated, visionary founder.

The Shape Already Exists

On April 2, the New York Times reported that Medvi, a telehealth seller of GLP-1 drugs run by Matthew Gallagher and his brother, booked $401 million in 2025 sales. The brothers are the only full-time employees, according to Drug Discovery & Development. CareValidate and OpenLoop Health carry the doctors, pharmacies and compliance.

Medvi is no model of conduct. The FDA sent it a warning letter in February, and reporters found its ads running under physician personas that appear fabricated.

Its structure is the lesson:

• One person orchestrates.
• A licensed institution signs.
• The customer already wants the outcome.

The nonprofit version is what I call owed-money work. An institution already owes a person something under its own rules or federal law, and the person loses it because the paperwork is hard. Nobody has to be persuaded to want the outcome. The money is already appropriated. The only thing missing is the paperwork, and paperwork is what agents do best.

Three markets pass that test at scale.

Path One: Medicaid Renewals

During the 2023–24 Medicaid unwinding, states disenrolled more than 25 million people, and 69% of them lost coverage for paperwork or procedural reasons, according to KFF. That is roughly 17 million people cut off over forms.

The next wave has a date. The 2025 reconciliation law requires states to redetermine eligibility for expansion adults every six months for renewals on or after December 31, 2026, and to impose work requirements by the same date. CBO estimates the work requirement alone will add 5.3 million people to the uninsured in 2034.

Arkansas showed what that looks like in 2018. More than 18,000 adults lost coverage under its work requirements. A Harvard team led by Benjamin Sommers found no change in employment and traced most losses to confusion and red tape. These are eligible people losing coverage to a missed form.

The value per case is enormous. MACPAC puts Medicaid spending at $8,677 per full-year newly eligible expansion adult in fiscal 2024. Keeping 115,000 eligible people covered for a year is $1 billion in care the government actually pays for. It is the cleanest billion in this whole field, counted in cash.

The hard part is reach. KFF lists outdated contact information as a main driver of procedural loss. The solo founder solves that with a partner that already holds the relationship: a community health center or a Medicaid health plan, both of which lose revenue when a patient loses coverage.

The founder brings the agents, and the partner brings the contact list.

Path Two: Hospital Charity Care

The Affordable Care Act requires every nonprofit hospital to adopt and widely publicize a written financial assistance policy. In 2019, Kaiser Health News found that 45% of nonprofit hospital organizations routinely billed patients whose incomes qualified them for free or discounted care. Dollar For estimates the gap at $7 billion.

Dollar For is already proving how far a lean team can go. Its 2025 Form 990 reports 18 employees, $1.64 million in expenses, 18,000 patients helped and $55 million in medical debt erased. That is $3 million in relief per employee.

In June, the OpenAI Foundation named Dollar For a grantee of its People-First AI Fund. Founder Jared Walker said AI puts Dollar For on a path to serve nine times more patients and eliminate $1 billion in medical debt.

That billion is face value at hospital list prices, and the hospital’s cost of the care is about a third of it. Even so, this is the furthest any organization has publicly pushed the idea, and it has a 990 behind it.

Path Three: Insurance Appeals

Medicare Advantage insurers denied 4.1 million prior authorization requests in 2024, according to KFF. Enrollees appealed 11.5% of those denials, and insurers overturned 81% of the appeals. About 3.6 million denials went uncontested.

The appeals that were filed were the strong ones, so the untouched pile will win less often. When Medicare’s independent reviewer sees cases nobody chose to push, it overturns about 5%. Even at that floor, the pool is worth $56 million to $156 million a year. On the services where reviewers reverse most often, 17% of PET scan denials and 15% of MRI denials, the agents would start where they win.

The market has noticed. Counterforce Health in Durham and Claimable both offer patients AI-drafted appeals, and both follow paying customers, clinics and drugmakers. Patients, who pay nothing, are still waiting for a nonprofit.

What Makes This Model Scale

Walter Kerr, a co-founder of Unlock Aid and a fellow at Renaissance Philanthropy, built and tested a free bill-dispute tool with real patients. He told me the tool works and has success stories, but wins came too rarely and ran too small against the size of the bills to build a scalable organization around.

His experience points to the design rule. A dispute over a few erroneous charges leaves the patient owing the rest. A charity-care approval erases the whole bill. An overturned denial covers the whole service. A renewal restores a full year of coverage. The solo models that can reach a billion contest an entitlement that covers the whole obligation.

The Leap of Faith

This is the part I can’t prove yet, but I believe in it.

No organization has published a retention rate for Medicaid renewal assistance. No agent-run shop has filed a 990 showing its cost per case. Benefits Data Trust, the Philadelphia nonprofit that helped 120,000 people claim $182 million in 2023, closed in 2024 after its call-center costs outran its funding. A one-person shop carries a new risk: it ends the week its founder burns out.

I’m making the bet anyway, for three reasons.

1. The cost structure is different. BDT’s costs moved with headcount. An agent-run shop’s costs move with compute, and a funding gap slows the work instead of ending it.
2. Two of the three paths skip BDT’s hardest problem. An appeal starts with a denial letter the patient already holds. A charity-care application starts with a bill the patient already holds. The person brings the data.
3. The arithmetic leaves room for error. Add the floors on appeals and charity care and one founder moves $390 million to $490 million in real resources a year, before Medicaid. Keeping 59,000 to 70,000 eligible people covered for a year, about half the 115,000 threshold, closes the gap to a billion. The founder can miss the full Medicaid threshold by tens of thousands of people and still get there.

Action Items for Funders

The founder exists somewhere already. What she lacks is a funder willing to back an org chart with one name on it. Four moves would change that:

• Fund a renewal pilot before December 31. Pair one solo builder with one community health center or Medicaid plan in a state with high procedural disenrollment, and measure retention against a comparison group. That one number unlocks the whole model.
• Fund the second operator. Write succession into the grant: an escrow of code and credentials and a trained backup, so the organization survives its founder’s worst month.
• Fund evidence. Pay for an honest test of whether upstream relief changes health and finances, so the next funder can underwrite with data instead of faith.
• Count in cash. Ask grantees to report the hospital cost of debt erased and the Medicaid dollars retained. That turns a face-value pledge into a result other funders can compare.

Amodei gave Silicon Valley a year. Medicaid’s new rules start in less than three months. If you are building this, or want to fund it, I want to talk to you.

Do you think a billion-dollar solo AI nonprofit is possible in your country too? Tell me how!

The post What Will the First $1 Billion Impact Solo AI NGO Look Like in Your Country? appeared first on ICTworks.

Last December I told Belal Jahjooh he was right about digital agriculture. I was right about the wrong half of it.

Jahjooh’s policy paper, Born Poor, Die Poor, argues that “digital solutions cannot compensate for unfair markets.” I went further in the comments and wrote that digital agriculture fails because it does not change a system designed to keep smallholders poor.

AgriTech founders and their funders made the opposite bet. In CTA and Dalberg’s Digitalisation of African Agriculture Report, advisory services held 22.6 million of 33 million farmer registrations, or 68%. Market linkages held 8%.

Jahjooh and I are right about the 68%. We are wrong about the 8%. Digital agriculture raised incomes where it put a second buyer in front of the farmer.

Sign Up Now for more digital agriculture insights

Advice Apps Move Income by Single Digits

A 2025 meta-analysis by Robert Beach and colleagues found that digital information interventions raised farmer income 6% on average, with a confidence interval of 2% to 9%. Five of the ten studies they used to assess income found no significant change. Raissa Fabregas, Michael Kremer, and Frank Schilbach argued in Science that gains this size still beat delivery costs by an order of magnitude.

I accept that. A cheap 6% is a decent extension program. The farmer still sells to the same trader on the same terms.

Price Information Needs a Second Buyer

Sandip Mitra and colleagues gave West Bengal potato farmers in randomly chosen villages the wholesale price. Average farm-gate prices did not move. The farmers bargained with village middlemen whose only competition was other middlemen. Knowing the price gave a farmer nowhere else to sell.

The Fishermen Fight Is About the Second Buyer

Robert Jensen’s 2007 Kerala study reported that after mobile coverage arrived, 30 to 40% of fishermen sold outside their local market and profits rose 8%. In 2016 I called it the largest lie in ICT4D. I was amplifying Jacques Steyn’s argument that licensing and auctioneer credit tied fishermen to one landing center.

Janaki Srinivasan and Jenna Burrell’s ethnography found Jensen’s results held in north Kerala, where geography and credit let fishers sell at different markets. In south Kerala, where credit tied them down, his results generally did not hold. Jensen named the same constraint in his own paper.

Jensen and his critics disagree about Kerala. They agree that the phone paid off only where the fisherman could sell to someone else.

Karnataka Shifted Power Until Traders Stopped It

Karnataka’s Unified Market Platform let outside traders bid in every state market. Retsef Levi and colleagues found in PNAS that it raised paddy prices 5%, which they calculated as profit gains of 36% to 159% on thin smallholder margins. One co-author worked for the platform’s operator.

For cotton, green gram, and tur, prices did not move.

The authors tied the difference partly to in-market concentration. Where a few traders controlled a crop, they kept controlling it on the platform. India’s agriculture ministry says the same thing plainly: states boost trade on e-NAM by licensing all registered buyers to bid across mandis. The lever is a licensing rule.

The Easy Version Is the One We Built

I have written the easy version myself. In May 2024 I wrote that AI brings hope to Nigerian smallholders, and I never asked who buys their crop.

Donors can fix this in a grant application. Require applicants to count the buyers a farmer can sell to today and after the project. If the count stays the same, fund it as extension and price it that way.

If a digital agriculture project cannot name the second buyer, it is extension with a login screen.

The post AgriTech That Raises Farmer Incomes? It Adds a Second Buyer appeared first on ICTworks.

In January 2012, Uganda’s Ministry of Health issued a moratorium halting new eHealth projects. Donors and NGOs had piloted up to 80 different mHealth applications in Uganda in 2008, and none of them scaled. Donors paid for the pilots. Nobody paid for national rollout.

£150,000to Scale Innovations

The 100x Impact Accelerator at the London School of Economics funds the stage after proof. It wants ventures with evidence that their model works and a credible route to hundreds of millions of people.

Each selected venture receives £150,000 in catalytic grant funding and a 16-week accelerator programme on scaling strategy and impact measurement.

The portfolio leans digital. Intelehealth is a 100x alumni. I can personally vouch for their support. Adalat AI, Darsel, Wysa and the Humanitarian OpenStreetMap Team are also in the portfolio.

Who Can Apply

• For-profit and non-profit ventures registered or operating anywhere in the world
• Ventures working in one of 100x’s eight impact areas, from health and social care to democracy
• Ventures with a track record of adoption and validated outcomes..
• 100x aims for 70% of its portfolio from emerging markets and favours founders based where the venture operates

Apply Now! Deadline is November 2, 2026

More Funding Opportunities

Please sign up now to get our email updates. Learn how to get startup funding for your technology business, and find new funding opportunities with donors.

• Do African Startup Bills Help Technology Firms?
• 4 Technology Startup Company Funding Sources for African Entrepreneurs
• 9 Ways Artificial Intelligence is Improving Nigeria’s Agriculture Outcomes with Smallholder Farmers

The post Apply Now: £150,000 for Digital Ventures Ready to Scale appeared first on ICTworks.

Large philanthropic money moves to organizations that can take large philanthropic money. Everything else about grant making is downstream of that.

In August, I wrote on ICTworks that the new AI philanthropies had not yet learned our habits and could still be persuaded to fund differently. The OpenAI Foundation sounds persuadable. On August 13, Anna Makanju wrote that organizations providing critical local services have historically been among the last to benefit from new technology, and that “we should not allow that pattern to repeat with AI.”

I had the mechanism wrong. Advocacy is not the binding constraint. Absorption is.

Sign Up Now for more AI philanthropy insights

First AI Grants Went to Capacitated Organizations

On September 10, the OpenAI Foundation committed $60 million to AI weather and crop-disease forecasting for 100 million smallholder farmers across South and Southeast Asia and East Africa over three years. Six organizations share it:

• University of Chicago,
• Precision Development,
• UC Berkeley,
• AIM for Scale (via University of Notre Dame),
• Digital Green,
• CIMMYT.

Three of the six agreements sit with American universities. Precision Development was co-founded by Michael Kremer, who also directs the University of Chicago’s Development Innovation Lab.

Every one of those six can take a multimillion-dollar wire this quarter. They hold the tax status, the audited financials, the indirect cost rates, the insurance, and the grants office that a funder’s legal team requires before money moves.

The health commitments show the same selection.

• In April, the Foundation committed more than $100 million across six US research institutions for Alzheimer’s.
• In August, it gave $100 million to the Common Health Coalition for AI that helps care teams reconnect hepatitis C patients who fell out of care, starting in Alabama, Illinois, Louisiana, and Massachusetts.
• In September, it launched Public Data for Health with more than $125 million.

Of roughly $435 million in OpenAI Foundation program commitments this year, $60 million targets low- and middle-income countries, and no named health grantee is based outside the United States.

The hepatitis C work is a community health workflow. Finding patients lost to follow-up and directing scarce staff time is what frontline programs in Kenya and Nepal have done on phones for decades. Those programs did not lose a competition. They were never in the pool, because the money needed a US counterparty.

The Pipe Is Older Than the AI Money

None of this started with OpenAI.

The Council on Foundations and Candid analyzed US foundation grants from 2016 to 2019 and found 39% of global grant dollars reached non-US recipients, with about 13% of dollars intended for international causes going to organizations registered where the work happens.

By my subtraction, about 61% stayed with US-based recipients.

Anyone who watched the last technology fortune enter global agriculture saw the same shape. When the advocacy group GRAIN analyzed that foundation’s food and agriculture grants from 2003 to 2020, it found almost half went to CGIAR, AGRA, the African Agricultural Technology Foundation, and large international bodies.

Of the money reaching hundreds of other recipients, 82% went to groups based in North America and Europe, and under 10% to groups based in Africa. GRAIN opposes that foundation’s approach, and the categories are theirs, so weigh it accordingly.

Two decades, two fortunes, one result. The constant is the receiving infrastructure.

Absorption Is a Capability, Not a Virtue

On August 20, the OpenAI Foundation said much of the organization is still left to build.

That means a program officer with $1 billion to move and a small team has a brutal arithmetic problem. A team in that position writes $10 million checks, because 100 checks of $100,000 need 100 due diligence files, 100 audits, and 100 reporting relationships.

Absorption means a grantee can take $10 million without breaking:

• legal standing the funder’s counsel recognizes,
• financials in a format the funder’s auditors read,
• the hiring pipeline to triple a team in 18 months,
• the governance to survive the scrutiny that follows.

A Nairobi-registered NGO running on $2 million a year may do better work and still fail that test on every line. That is a gap in sector infrastructure, and infrastructure can be built.

Coefficient Giving Is Building Capacity on Purpose

Coefficient Giving named the problem and is funding the answer.

On September 8, Otis Reid and James Snowden wrote that the organization is accelerating its global health spending to build capacity for a much larger future wave, because the best opportunities cannot scale overnight and organizations need time to hire, expand, and build infrastructure.

The instrument is Launchpad Funds, a request for proposals that offers $200,000 to $10 million in pre-seed money to develop a new fund, and $10 million to $50 million or more to launch one at scale. Coefficient says it weights each proposal by its potential to absorb $100 million or more from other funders over the next several years.

That is a grantmaker paying to create organizations that can receive money it has not yet raised.

The same logic runs through the $165 million Health Aid Transition Fund, announced September 18, which builds on about $18 million in earlier grants. One stated aim is helping governments select local partners to take over from international contractors, alongside work on budget absorption, where the fund’s own materials note that low- and middle-income countries underspend health budgets by an average of 13%.

The open question is who ends up leading the funds this seeds. The RFP sets no geographic limit I could find, and the pre-seed track is small enough for an organization outside the US to pursue. If the general managers all turn out to be in Boston and London, the capacity gets built in the same place it already sits.

What the Sector Should Build Now

I would like to see three things:

1. Pooled funders capitalize regional receiving funds, registered and governed in Nairobi, Delhi, Dakar, or Jakarta, each run by a manager accountable for an outcome and sized to absorb $100 million. Coefficient’s pre-seed track pays for exactly this kind of design work, and nobody has to invent the vehicle from scratch.
2. Intermediaries already holding large grants publishing the share of each grant that reaches locally registered organizations and the date it arrived. That number is the sector’s only honest measure of whether the pipe is changing.
3. Funders writing new AI eligibility rules stating plainly whether equivalency determination is available and what it costs an applicant, so an organization in Kathmandu learns in two minutes whether applying is worth two weeks.

The next wave of AI money will go wherever it can be absorbed in 90 days. Build the thing that can take it.

The post We Have Not Yet Built The Capacity to Absorb AI Philanthropy appeared first on ICTworks.

On 15 September, Kirsty McNeill sat for her first scrutiny session as Minister for International Development and Africa. Across 78 questions, nobody in the room said digital development, connectivity, or digital public infrastructure. Not the minister, not the permanent under-secretary, not one MP.

Three weeks ago I argued that FCDO had a published 2030 digital commitment and no visible money behind it, and treated that as a budget problem waiting for a budget fix. The transcript of McNeill’s session says something harder.

Digital development is being reclassified out of development, and there is a worked precedent inside the same department for where reclassified sectors end up.

Sign Up Now for more digital development funding analysis

78 Questions on International Development

McNeill’s oral evidence session ran through Africa, Ukraine, Gaza, Sudan, Ebola, water, education, climate finance, women and girls, children, civil society, communities of expertise, the FCDO restructure, ICAI, and relations with the EU and the US. Nick Dyer, the interim permanent under-secretary, sat alongside her.

The only thing in the digital family that arose came from McNeill in passing. She told the committee the AI Minister is aware of both the potential of AI for development and its risks, and already wants to work with her on it. AI for development is now an offer arriving from another department.

Two answers settled questions that were open when I last wrote.

• On the March allocations, McNeill told the committee they are not up for being reopened, which fixes the Global Tech thematic directorate at £24 million for 2026/27 against total FCDO programme ODA of £6,815 million.
• On direction, she confirmed continuity with Baroness Chapman’s Africa Approach and with the four shifts.

She also gave the government’s answer on Africa. Asked whether the bilateral cuts fall disproportionately on the continent, McNeill said FCDO expects the proportion reaching Africa to hold steady once imputed multilateral spend is counted.

That is a real answer on the Africa total. It is no answer at all on the four digital targets, which no multilateral has agreed to deliver on the UK’s behalf.

Absence from one transcript is weak evidence by itself. The committee sets its own agenda and had a great deal to get through. What gives it weight is that this is the body that checks whether FCDO delivers published commitments, and FCDO has four published digital commitments due in 2030. Nobody asked about any of them.

WASH Went Away First

Brian Mathew MP raised water, describing UK backing for WASH as decimated. Dyer’s reply is the most useful passage in the transcript for anyone working in digital development.

He said that around six years ago FCDO moved away from direct service delivery in water towards strengthening systems and leveraging finance. Then: “We are unapologetic about that.” He went on to describe the World Bank’s Water Forward commitment as where the scale now sits.

Read the sequence.

1. Direct service delivery becomes system support.
2. Grants become expertise.
3. Money gets leveraged through multilaterals.

A decade later an MP tells the department a sector the public actually understands has been gutted, and the department answers that the World Bank is doing it now, and says it would make the same choice again.

Digital development is at the start of that sequence.

Two of the four shifts in the July 2026 modern development approach are service delivery to system support and grants to expertise, in those words. The Digital Development Strategy already routed digital infrastructure through British International Investment and Mobilist before any of this, which we noted at its launch without seeing where it led.

Dyer also told the committee the research programme has been cut by 50%, and that FCDO has shut some of its centrally managed programmes. He did not name the Digital Access Programme, which is centrally managed.

My prediction: within five years, FCDO’s answer on connectivity will be the one it now gives on water, that the World Bank and private capital are handling it.

Digital Went Two Places

Here are three ministers, in one month, issuing three framings of digital technology as:

• A UK export capability.
• A frontier safety problem.
• A cause of loneliness in Britain.

None of them is about a person in a low-income country getting online. So where did digital go?

Trade, at post level.

On 2 September, FCDO commissioned the Siun Community Self-Built Connectivity Hub in Ogun State, one of ten community hubs under the Digital Access Programme, implemented by the Initiative for Digital Inclusion. The FCDO official who spoke was Grace Bell, First Secretary for Economic Partnership. The press statement came from FCDO’s communications lead for Growth, Trade and Investment in Nigeria.

The same week, McNeill opened the UK-Angola Investment Forum and described what the UK brings as “expertise in finance, infrastructure, technology and innovation.” The infrastructure she then listed as already delivered: hospitals, roads, electricity networks, flood defences, and the Lobito Corridor railway. No connectivity.

This was set up in December 2025, when Chapman launched the Africa Approach with seven principles and put digital in the seventh, alongside “music, sport, and design.” In the same speech she backed Mission 300 to reach 300 million people with electricity by 2030. Electricity got a number and a date. Internet access got a policy training programme for diplomats.

Security, at Cabinet level.

On 19 September, McNeill told an audience at SOAS that the digital revolution is “making us sick and it is making us sad”. Then on 22 September, Andy Burnham told the UN General Assembly the UK will put AI at the heart of its G20 presidency, with the leaders’ summit in Manchester in November 2027.

The next day Ed Miliband addressed the Security Council on AI, setting out safety testing, transparency from AI companies, and resilience, and citing the AI Security Institute. The harms he named were jobs, child protection, cyber security, biosecurity, national security and financial stability.

The ICT4D Efforts Left Out

The GSMA, with FCDO funding, reports that 96% of the world lives inside mobile broadband coverage while 3.1 billion people still do not use it. Handset cost, digital skills, and content people can actually use are what hold that gap open.

Neither frontier AI safety standards, nor investment partnerships close it, because operators have already priced those users and walked away. The four DDS targets sit exactly in the space between the two agendas the UK has chosen.

The Digital Access Programme is still working that space. Its programme record runs to 31 March 2030 and hubs are operating in Kaduna, Kwara, Abia and Ogun. The delivery continues. What has gone is anyone in London who is accountable for whether it adds up to 20 countries.

My Four Predictions for FCDO

Three weeks ago I predicted the UK’s G20 presidency would centre on debt architecture. Nope. Burnham announced an AI agenda that contains no connectivity, no DPI, and no low-income-country capability.

Here are four more predictions I make with a heavy heart:

1. The Digital Development Strategy will not be withdrawn or replaced. It will reach 2030 with no progress report published against its four targets.
2. No Digital Access Programme successor phase will be announced from London. Hub openings will continue to be announced by High Commissions.
3. Connectivity and DPI will appear in UK G20 workstreams as adjacent to the AI standards agenda, with no ODA allocation attached.
4. Within five years, FCDO’s public position on connectivity will match its current position on water.

I only wish someone had asked a seventy-ninth question: what is to become of FCDO’s digital development efforts now?

The post FCDO Answered 78 Questions. Digital Development Came Up Zero Times. appeared first on ICTworks.