“Humans are involved in AI preparation, AI verification and AI impersonation,” Gemma Newlands unfolds.

How artificial is artificial intelligence, really?

We often think that artificial intelligence (AI) will make humans redundant, but is it really so? In reality, humans play a bigger role in AI development and deployment than most think.

The societal imagination of AI is fueled by fantasy more than reality. Very few of us are actually informed about the real value chain of AI.

Lifting the curtain of AI-as-a-Service

AI-as-a-Service means that organizations can access specific AI capabilities via the Cloud instead of dedicating in-house resources to AI development and maintenance. This kind of service is currently offered by major vendors including Amazon Web Services, Microsoft Azure Google Cloud, Oracle and Sales Force. AI-as-a-Service is also the key business model of many start-ups and small-medium enterprises. The services can range from conversational bots to knowledge mapping, computer vision and speech recognition.

The market for AI-as-a-Service has grown hand in hand with companies’ possibilities to gather and process enormous amounts of data

Gemma Newlands, a PhD Researcher at The Nordic Centre for Internet and Society wanted to investigate whether organizations live up to the ongoing techno-utopian narratives of AI hype. Based on existing literature and case studies Newlands was able to critically analyze the human labour involved in AI production and the lack of transparency surrounding it. What did she uncover?

Hidden human labour

Without a doubt, not all actions required to provide AI-as-a-service are artificial.

“Humans are involved in AI preparation, AI verification and AI impersonation. For instance, many do not know that for home assistants, such as Amazon Echo or Google Home, contracted workers review the effectiveness of AI by listening to audio recordings of conversations,” Newlands unfolds.

Humans are involved in several steps along the AI-as-a-Service value chain. In many cases organisational clients must also provide additional labour to adapt and make the AI services work. The AI vendor also provides human labour for AI co-production.

Strategic secrecy

So, is the mystery around artificial intelligence deliberate?

A benefit of keeping the human effort in AI production invisible backstage is to keep it coherent with the utopian techno narrative. Vendors promote the idea that AI can both mimic and surpass human intelligence and, in that way, make life easier. This promotion corresponds with the AI hype of reducing labour costs.

With this invisibility, important conversations about privacy and ethics among the users of AI-as-a-service are also avoided. Overall, the human effort is kept hidden to generate increased funding and avoid uncomfortable questions.

Why we should care

“Data work in AI production is usually performed in low-income countries and poorly compensated. The invisibility of the strategic work involved in AI production blurs the acknowledgment of human and technical achievement,” Newlands explains.

The strategic secrecy of human effort in AI-as-a-Service is problematic in terms of not ensuring transparency and accountability for users. Furthermore, the strategic secrecy may lead to dissatisfied customers or come in trouble with consumer protection agencies. Finally, another concern is how work is hidden in the global value chain, which may be detrimental for prestige, salary and recognition of this work.


Gemma Newlands: Lifting the curtain: Strategic visibility of human labour in AI-as-a-Service. Big Data & Society, 2021.

Powered by Labrador CMS