One AI impact company’s pursuit of proof

Sama, a San Francisco-based data-training company, decided a few years ago that it was time to run its impact numbers through an independent randomly controlled trial. The risk paid off.

Sama, which provides analysis and high-quality ‘training’ data to develop accurate machine learning models, saw outsourcing to countries such as Kenya and Uganda as a vote of confidence in the region. Done well, it could be something mutually beneficial. Metamorworks / iStock

In short

  • Sama provides data and content management services to businesses in a range of sectors.
  • It turned to the Massachusetts Institute of Technology (MIT) for independent proof that its impact model was working.
  • Empowering its workers at a bespoke university, Sama sees outsourcing as a vote of confidence in poorer regions.
  • Over half of its agents in Africa are women earning 60% higher than the average.

Sama provides analysis and high-quality ‘training’ data to develop accurate machine learning models. It specialises in image, video, language, and sensor data annotation and validation for machine learning algorithms in various industries. It counts Google, GM and Walmart among its clients.

Starting out as a non-profit with a clear impact-led vision, Sama has always been upfront about its people-over-profit ethos. It has also insisted that humans need to be involved in the data training process to keep biases out of the algorithms that run the show.

But as a business having a ‘human in the loop’ can mean one thing: higher production costs, lower profits. A purely profit-seeking business is primed to farm out such work to low-cost data centres. These are typically located in countries with cheaper labour.

For companies mindful of their impact (and their reputation) this decision can be problematic. Stories abound about exploitation in outsourcing contracts to the developing world.

Forbes’ advice on the sticky topic of how to bake impact into AI business favours keeping jobs closer to home with better training and employment conditions to “support employees’ long-term growth and success while helping to break down the tech industry’s oppressive barriers to entry.” That sounds expensive, too.

Sama looked for a third way. It wondered how struggling regions can pull themselves out of poverty if companies avoid them altogether. Sama saw outsourcing to countries such as Kenya and Uganda as a vote of confidence in the region. Done well, it could be something mutually beneficial.

Independent proof

Sama’s own data-gathering efforts appeared to back this thesis up, but with business models and reputations at stake it became clear that independent proof would still be needed.

So the company turned to a third party, Massachusetts Institute of Technology (MIT), who carried out a randomised control trial (RCT), arguably the gold standard for establishing real impact.

The decision was not without risks and costs. What if the impact couldn’t be pinned down or it was really too soon to tell? How would investors and the public interpret the published results?

Striking impact gold

Any nervous moments for Sama during the three-year study turned out to be baseless. The findings reaffirmed Sama’s belief that its sourcing model gives a boost to individuals but also their families and communities.

Sama reports that its activities have had a positive impact on over 55,000 people (and counting) in underserved regions.
Training and upskilling at Sama University (SamaU) have opened new doors to long-term careers in technology both within and outside the company.

People trained and given the opportunity to work at Sama report a striking 40% gain in real earnings compared to the RCT control group, and unemployment rates are 10% lower.

Founded by the late Leila Janah in 2008 and today run by Wendy Gonzalez, Sama’s focus on gender equality in the tech industry also stands up in the study’s findings. More than half of its agents in Africa are women whose average wages are 60% higher than the control group.

“Receiving results from our RCT study has been an amazing experience because I know Leila would be proud to know her mission to ‘give work’ is empowering thousands of women globally,” Gonzalez remarked in an earlier interview, “and the data to back this up is better than we could have ever imagined.”

Investor chuffed to bits, too

As an investor in the company since February 2019, Rubio Impact Ventures is clearly chuffed that Sama goes above and beyond to substantiate its impact-led vision for helping its people and the communities they live in.

“Sama is a fantastic showcase for our investment thesis that you can build great companies while growing business, growing impact and generating healthy returns that even compete on a global scale with leading artificial and machine-learning companies,” Warner Philips, a Managing Partner at Rubio told Impact Investor.

“We’d known Samasource, as it was then called, for several years and had always been impressed with its business and credentials, especially its world-class impact reporting,” says Philips.

“As a non-profit, most investors wouldn’t even look at it,” he adds, “and many tech investors were uncomfortable with the thousands of people working for them at the time in Africa and India.”

Sama weathered a very tough 2020, losing its founder and CEO while dealing with Covid-19 and putting together a complex work-from-home programme for 1,700 of its workforce in Kenya and Uganda.

It managed all this, Philips says, while building the business, growing the team, and developing new technology to help cement and expand its leading position.

Philips says he’s impressed with Sama’s commitment to continuous improvement in line with its core values. He thinks it gives the company a “differentiated position in the market,” making it more attractive to tech giants in several industries who are keen to leverage AI’s potential.

The effort seems to be paying off, as Sama recently earned a spot on Forbes’ top-50 most promising AI companies to watch in 2021.

Read here the rest of our interview with Philips.