The Fuzu Story: The AI Pivot That Beat The Odds
Nairobi, twenty ten. As Jussi Hinkkanen tells it, Nokia was hosting a conference in the city, and it needed one driver to chauffeur its VIP guests.
It advertised the job. Fifteen hundred applications came back.
The organisers agonised over how to choose. Fifteen hundred people for one seat behind one steering wheel, and no way to tell them apart.
Ten years later, in March twenty twenty, the company Hinkkanen built to sort people like them reported one and a half million active users.
In the same interview, he gave a second number. Five hundred thousand dollars of revenue.
Put the two side by side, and it comes to about a third of a dollar for every active user, over a year.
The company had done what it set out to do, and gathered the people. It could not get much money out of them, or out of the employers who hired them.
So what does a company do with a million and a half people who cannot pay it?
For ten years, the answer was patience. Then it found somebody else who would pay.
FIFTEEN HUNDRED FOR ONE
In twenty ten, Jussi Hinkkanen was a Finn who had been working across Africa and the Middle East since two thousand and four. He had lived in Mozambique, in South Africa and in the Emirates.
By then he was a vice president at Nokia, in charge of corporate affairs for Africa, India and the Middle East.
In the same city was another Finn from the same company. Jussi Impiö was the founding head of Nokia’s research centre for Africa, in Nairobi.
Every later version of the driver story traces back to a single interview Hinkkanen gave in twenty fifteen, so it stands as his account.
The problem inside it is plain enough. It was not that nobody could do the job. It was that too many people could, and there was nothing to tell them apart.
How do you choose one person out of fifteen hundred when you cannot possibly meet them all? The difficulty, as that interview put it, planted an idea that took five years to become a company.
One person got the seat. Fourteen hundred and ninety nine did not. Who they were is not recorded. They are simply the shape of the pool this story is about.
In twenty thirteen, Hinkkanen quit Nokia. On the twenty fifth of June that year, a company called Fuzu was registered in Helsinki, with Hinkkanen and Impiö as co-founders.
It was Finnish on paper and aimed at Nairobi. Its startup costs ran to as much as ten million Kenyan shillings.
That money came from shareholders, from the Rockefeller Foundation, and from a Finnish public funding institution. Finnpartnership, a Finnish public support programme, backed it through twenty fourteen and twenty fifteen.
A great deal of the money behind Fuzu over the next decade would come, one way or another, from the Finnish state.
On the seventeenth of September, twenty fifteen, Fuzu launched in Nairobi. It had about fifteen staff, and a third of them were Kenyan. The two founders held a majority among about ten shareholders.
It had also hired Robert Kimani, who had been chief executive of BrighterMonday, the established Kenyan job board, as its president for business development.
The idea was to do for every employer what the conference organisers could not do for one driver. Test the applicants, rank them, and put the strongest at the top of the pile.
The model charged both sides from the first day. Employers got their first job post free. Testing and ranking candidates cost fifteen thousand shillings, and access to CVs about five thousand shillings a month.
Job seekers could upload a CV for free. To see how they ranked against the other candidates for similar jobs, they paid about two hundred and ninety nine shillings, once.
Or they could pay about nine hundred and ninety nine shillings a month.
So the people with the problem, too many applicants and too few jobs, were customers too. Plenty of job boards charge this way. Fuzu did it from the start.
Hinkkanen described the ambition as turning, in his words, hopeless job hunting into an inspiring journey of self discovery.
And people came. Within forty days of launch, thirty thousand Kenyans had signed up.
Thirty thousand, in a little over a month. The pool had begun.
THE SORTING MACHINE
A job seeker on Fuzu built a profile, then took psychometric tests and skills tests. They built a CV with the platform’s tools, and they could take its courses.
Every test a job seeker took told the platform something about them. Every new profile made the pool bigger.
On the other side, employers posted jobs and got ranked candidates back. An employer facing a pile like the conference organisers’ could now see it sorted.
For a job seeker, the tests were a way to be seen. For an employer, they were a way to cut a pile down.
In November twenty sixteen, Finnfund took equity in Fuzu. Finnfund is the Finnish state’s development financier. It put in just under one million euros.
The round was announced that December at one point seven five million euros, with the Finnish staffing group Barona and an investor called Polkuni alongside.
In June twenty seventeen, Finnfund published an article about the company it had backed, and it put Kenya’s problem in numbers.
Forty six million people. Eighteen million of them economically active. Three million in formal employment.
And about fifteen million Kenyans unemployed or underemployed.
Fifteen million people was the size of the pile the machine was being built to sort.
The article also described bribes paid for interviews, and jobs that were, in its words, predetermined before interviews.
So a machine for sorting applicants was being built in a market where, by that account, some jobs were settled before the sorting even began.
In that article, Finnfund put Fuzu’s Kenyan users at more than eight hundred thousand. The company had twenty one staff.
And job seekers were still paying. For a hundred shillings, a user could buy something Fuzu called Instant Feedback.
A hundred shillings is a small sum. But the people paying it were people looking for work. The pool was growing faster than the money.
In twenty eighteen, Fuzu launched in Uganda.
The Finnish public support kept coming. Business Finland, the state’s innovation funder, backed Fuzu’s research and development more than once, most recently in twenty twenty.
This was patient money. A development financier does not need a quick exit, and that patience is part of how a company like this keeps going.
By his own account, Hinkkanen was the primary architect behind the platform. He said he spent half his time on the ground, two weeks of every month.
On the tenth of March, twenty twenty, Fuzu closed a Series A of three point four million euros, led by Sparkmind.
Finnfund came in again, alongside Cornerstone Enterprises, Aucfan Incubate, Seedstars International, Kepple Africa Ventures and Barona. The money was for artificial intelligence in the recruitment product, and for expansion into Nigeria.
It is the largest round Fuzu has ever raised.
After five years of watching the machine being built, investors believed in it.
A THIRD OF A DOLLAR
On the day of that Series A, Hinkkanen gave an interview. In it, he said Fuzu had one and a half million active users over the previous six months.
He said the company had about five hundred thousand dollars of revenue. It had thirty five staff and was aiming for sixty.
Put those numbers together, and a platform with one and a half million active users was bringing in roughly a third of a dollar for each of them, for a year.
That is an order of magnitude, not an exact figure. And it did not come from a critic. It came from the founder, on the day of his biggest raise.
By his count, Fuzu had raised about five million dollars since twenty fifteen. Five years in, a year of revenue came to about a tenth of the money that had been put in.
It measures more than one company. It is what African employers were willing to pay to hire, and what people without work could afford to pay to be seen.
The filed accounts that followed said much the same. Fuzu’s Finnish parent company reported revenue of five hundred and ten thousand euros in twenty twenty one.
In twenty twenty two, the Finnish parent company reported five hundred and fifty nine thousand euros.
Those are the parent company’s own accounts. Whether they include the Kenyan, Ugandan and Nigerian companies, the public filings do not show.
Twenty twenty two was not a quiet year. In December twenty twenty one, Fuzu had launched in Nigeria, in Lagos, on the money from its largest round.
By twenty twenty two it had eighty five staff.
And revenue at the Finnish parent company grew by nine point six percent.
Its operating losses were one point three eight million euros in twenty twenty one, and one point one two million euros in twenty twenty two. Together, that is about two and a half million euros in two years.
That was not a collapse, and nobody on the record called it one. It was a squeeze, paid for by the people who had put money in.
Employers did not pay more because they did not have to. Jobs were shared for free in WhatsApp groups.
Job scams had taught people to distrust listings. And most work was informal.
BrighterMonday’s parent company, owner of the established job board, had already shut some of its African markets in twenty sixteen, and in twenty twenty two it regrouped under a new name.
Fuzu was still there. But it survived partly by staying small, and partly on funders who were not waiting for a quick exit. Surviving did not mean its model had beaten anyone.
Year after year the losses built up, until they outran everything the shareholders had put in. The Finnish parent company has negative equity.
Through all of this, Fuzu published numbers about its users. Registrations. Active users. Profiles.
A Finnfund project page in twenty twenty two quoted about three hundred and fifty thousand monthly visitors, and more than five thousand applications a day.
What Fuzu has never published is how many people got a job through it. How many job seekers paid, and whether paying helped them, is not public either.
The funders stayed. In September twenty twenty two, Finnfund and Sparkmind put in a mezzanine facility of one point two million euros. Finnfund’s share was five hundred thousand euros, and Sparkmind put in the rest.
A mezzanine is closer to a loan than to a new round of shares. Nobody put a fresh price on the company.
And Finnfund wrote down what the money was for. B to B operations growth. Business to business.
In twenty twenty two, on the funders’ own paperwork, the money was already pointing toward business customers.
A DIFFERENT BUYER
By about twenty twenty three, Fuzu was selling something new. It was called Fuzu Global Workforce.
A company in Europe could hire a remote employee in Africa without setting up a subsidiary there. Fuzu handled the payroll, the tax and the HR.
In the trade, this is called an employer of record. The European company directs the work, and the service carries the employment.
For a job seeker in Nairobi, it could mean a remote job with a company in Europe, with Fuzu handling the pay.
Global Workforce marketed a pool of two million people. About ninety percent of them were in Kenya, Uganda and Nigeria, with others in India and Pakistan.
The people were the same people. The job board that had gathered them was the same job board. What had changed was who was paying, and for what.
Those two million had come in through the same front door as everyone else. A free CV upload, and a set of tests.
The pool had been built by asking people to prove what they could do. Tests, profiles, CVs. For a remote employer, that was the product.
For years, employers had paid Fuzu to find someone among its job seekers. Now a buyer was paying for the job seeker’s work.
The twenty twenty two paperwork had said business to business. This was what that looked like.
The Finnish parent company’s revenue rose to eight hundred and thirty two thousand euros in twenty twenty three, up forty nine percent.
In twenty twenty four, the Finnish parent company’s revenue reached one point four three million euros, up seventy two percent.
The accounts do not say where that growth came from. They do not split the job board from the remote hiring business.
Revenue that had barely moved in twenty twenty two climbed in the years Fuzu built remote hiring. The timing is what the record shows.
The losses narrowed as well. From one point three eight million euros in twenty twenty one, the operating loss fell to seven hundred and one thousand euros in twenty twenty three. Every earlier year in the public accounts had ended in a loss.
In twenty twenty four, the Finnish parent company made its first operating profit. One hundred and thirteen thousand euros. Eleven years after it was registered.
In January twenty twenty five, Hinkkanen posted about something he called L L M Coding Gurus at Fuzu Remote. The title of that post is all the public record holds of it.
The same year, Fuzu launched its first fully white labelled platform in Europe. Its software was now running under another company’s name. Another business could run a hiring platform on what Fuzu had built.
And the company that owns all of this is small. The Finnish parent, registered in Helsinki, has three employees.
It is a small head office in the north, on top of a business whose people are in Nairobi, Kampala and Lagos.
In twenty twenty five, the Finnish parent company reported revenue of two point two million euros, up fifty five percent. It is the best year on file.
The operating result was a loss of sixty seven thousand euros, close to break even. Its equity was still negative.
THE INTAKE
Type the old Global Workforce web address today and it forwards you somewhere else. It lands on Fuzu Atlas.
Fuzu describes Atlas as governed AI data operations. It sells the human work that goes into training and testing artificial intelligence.
Annotation for reinforcement learning from human feedback. Evaluation of large language models. Safety red teaming, where people try to make a model misbehave.
Multilingual data in Swahili, Hausa and Amharic. It markets more than three million validated profiles.
And in its own words, it sells a dedicated, managed team operating on an ongoing basis.
The asset is the one Fuzu spent a decade gathering. The buyer is new.
Early in twenty twenty six, Jobtech Alliance invested in Fuzu. The amount was not disclosed.
The news reached the press that July. The deal came with a six month engagement to help Atlas grow its international clients.
The public account of what Fuzu had become came from Jobtech Alliance’s investment lead, Nick Markham. Most African talent platforms, he said, still operate as marketplaces, competing on volume and cheaper sourcing.
Fuzu, he said, has already moved past that. They have built the infrastructure layer underneath the marketplace.
That is the investor telling the story. In any source from twenty twenty six that we could find, the founder has not told it.
Jobtech Alliance also described the pressure behind the change. Increasing competition. Tightening margins. And clients who wanted managed teams rather than access to a pool of talent.
Set that next to a third of a dollar per active user, and Atlas does not look like a company chasing a fashion. It looks like arithmetic.
As job seekers, the people in the pool could barely be charged. As workers, there were buyers for what they could do.
Atlas recruits on fuzu dot com, in the same listings where people look for ordinary jobs.
One posting is for an English language transcriptionist, at four hundred shillings an hour.
Another asks people to film themselves doing everyday activities.
It says payment is made only for videos that meet the required quality standards and are accepted by the team.
A third asks English speakers to talk for a minute at a time on set prompts, three recordings a day for four days. Atlas recruits outside Africa too, including Czech transcribers.
The person scrolling that board for a job is the person Atlas is recruiting.
Jobtech Alliance describes Atlas as real oversight and accountability, rather than anonymous task based work. It talks about quality jobs for the people doing the work, particularly women in digital work.
The postings describe freelance work, paid by the hour or by the accepted video. Both pictures are current.
The public record does not say which of the two is the bigger part of Atlas.
Nor does it say who is buying. Fuzu does not name its AI clients.
On the seventh of March, twenty twenty four, Remotasks cut off its Kenyan workers. It stopped operating in the country the next day.
In April twenty twenty six, Sama issued redundancy notices to one thousand one hundred and eight workers in Nairobi, after Meta ended a content and data annotation contract.
Neither of those happened to Fuzu, and neither is a forecast. They are what dependence on a few foreign clients has already looked like, in the same city and the same kind of work.
Because Fuzu does not name its clients, how much Atlas depends on a few of them cannot be judged from outside.
What the people doing Atlas work earn over a month, and what they think of it, nobody has yet asked in public.
And were the people who built their profiles to find a job told that the same pool now supplies AI data work? The record does not say. Recruiting from your own users can be entirely lawful.
In June twenty seventeen, a Finnish development fund published the names of two Kenyans looking for work through Fuzu. Gregory Oyolo. Sylvia Biwott.
In the public record, they are the only people who used Fuzu to look for a job and are named.
It began with fifteen hundred applications for one driver’s seat. Fuzu’s founding line is that every human being is born with talent. For ten years it tried to help people with talent and no job get found, and it could barely charge them for it.
Now some of the people who came to Fuzu looking for work are what Fuzu sells. Whether that fulfils the line or reverses it, the record cannot yet say. This is Asili Africa. Every empire has an origin. See you on the next one.
Key Takeaways
- FIFTEEN HUNDRED FOR ONE. In twenty ten, Jussi Hinkkanen was a Finn who had been working across Africa and the Middle East since two thousand and four.
- A THIRD OF A DOLLAR. On the day of that Series A, Hinkkanen gave an interview.
- A DIFFERENT BUYER. By about twenty twenty three, Fuzu was selling something new.
- THE INTAKE. Type the old Global Workforce web address today and it forwards you somewhere else.
Also available on YouTube — search “Asili Africa” or subscribe to our channel.