AI in the Workplace- Achieving Efficiency Without Harm
A computer can never be held accountableTherefore a computer must never make a management decision
- Attributed to an IBM training manual, 1979
In October 2018, Reuters, the British news agency, reported that Amazon had silently binned its AI recruiting project, which it had been developing for more than three years. The reason? The AI was unfairly biased towards recommending male candidates and penalising applications that indicated that the candidate was a woman. The system was trained on resumes submitted to the company over the past 10 years. Since, historically speaking, men dominated the tech industry; the system ended up learning from a largely male applicant pool and started favouring these patterns, leading to a bias against women. Amazon ended up discarding this project in late 2017 after finding that the system could not be reliably made gender-neutral.
The AI was doing exactly what it was trained to do: identify the patterns in the past data to make decisions for the future.
But what happens when the patterns themselves reflect bias?
Eight years later, we see that the discourse has remained, if not become even more, relevant in today's landscape where organisations have started using AI in various functions like hiring, performance evaluations, monitoring and even managing employees.
At the same time, we must not overlook the fact that this technology has improved by leaps and bounds over the years. AI today promises faster hiring, improved productivity, efficient decision-making, and reduced workloads. It has automated repetitive tasks that used to weigh employees down and helped streamline processes, enabling them to shift their focus on tasks that require human judgement and creativity.
To sum it all up, improving efficiency has become one of the major contributions of AI in the workplace. But the question stands: when technology starts influencing decision-making involving humans, should efficiency be the only metric used to judge success?
The answer becomes even more complicated when we consider the consequences of those decisions. Does an AI system have the full capability to judge, evaluate and monitor an employee’s behaviour? And if it does, how far can it go? Where do we draw the line between improving productivity and protecting privacy? What effects would it have on the employees’ sense of autonomy and dignity in the workplace?
And perhaps the biggest question of all:
How can organisations achieve efficiency while upholding fairness, human dignity and privacy?
The answer begins with one simple principle: Do No Harm.
Do No Harm, expressed differently across various frameworks, has become one of the foundational principles in the ethical adoption and use of AI. In a workplace context, this translates into looking beyond the efficiency achieved and focusing on whether the AI’s use is justified, proportionate and appropriate for the people affected, in particular, the employees.
This principle is reflected in UNESCO's Recommendation on the Ethics of Artificial Intelligence, adopted by its 193 Member States in November 2021. UNESCO identifies “Proportionality and Do No Harm” as its first core principle and it states that the use of AI must not go beyond the intended purpose. It also advises that risk assessment be conducted to identify and prevent any potential harm.
Consider AI-powered employee monitoring systems. Though they may be able to analyse and find inefficiencies in workflow, they also raise questions about employee privacy, autonomy, and dignity. Is the introduction of this system justified? Is the level of monitoring proportionate to the aim of introducing such a system? Are the employees informed about what data is being collected and used? And are there any safeguards placed within the organisation to prevent the technology from being misused?
Thus, this principle encourages organisations to determine whether the introduction and use of AI in each specific context is justified, appropriate, and responsible.
UNESCO’s principles also highlight the importance of fairness and non-discrimination, the right to privacy and data protection, transparency and explainability, human oversight, sustainability, and responsibility & accountability. UNESCO’s recommendations cover 11 policy areas which collectively address the issues that arise with AI adoption and use.
It also emphasises an important distinction:
Human involvement does not automatically translate to meaningful oversight.
While AI can assist in reducing workload and providing valuable inputs for decision-making, the consequences of these decisions cannot be transferred to the system. UNESCO clarifies that humans should not be replaced by AI in areas of responsibility and accountability.
The focus of this discourse, therefore, should not just be limited to whether a human ends up making the decision despite the source of information. It should also consider whether the decision maker has the capability to understand, analyse, and question the reasoning behind the recommendations made by the AI.
India’s Approach to Responsible AI
With global attention being directed towards AI, India has taken a stride towards Governance of AI by releasing a principle-based AI Governance framework in February 2026. The framework focuses on “AI For All”, aiming to ensure that AI is not concentrated in a handful of firms or geographies, but diffused across agriculture, healthcare, education, governance, manufacturing, and climate action. It is divided into four parts, the first of which lays down the foundations by setting out the seven principles that ground India’s governance philosophy on AI. These principles, also known as the 7 Sutras, include trust is the Foundation, People First, Fairness and Equity, Innovation over Restraint, Accountability, Understandable by Design, and Safety, Resilience & Sustainability.
These Sutras can become the guiding framework for organisations that have already integrated AI or are considering its adoption in their workplaces.
The People First Sutra emphasises the need to place people in the centre, with the deployment of AI contributing towards the strengthening of human agency. The principle also covers capacity building, ethical protections and safety considerations. Similarly, Fairness and Equity highlight the need to design AI systems through methods that reduce the risk of bias and discrimination.
The framework also calls for AI developers to be accountable for their respective roles in the AI-value chain. This places the responsibility not only on the users but also the people behind the designing of the system.
Altogether, the Seven Sutras highlight one important takeaway: Just having a human in the loop making final decisions is not enough.
From Efficiency to Responsible Adoption
The OECD and International Labour Organization's Compendium of Best Practices for a Human-Centred Development and Use of Artificial Intelligence in the World of Work further reinforce the importance of placing the employees and workers at the center. It suggests looking at AI through the lens of both workers and workplaces.
A human-centered approach requires organisations to look beyond productivity and efficiency, and take into serious consideration how the AI system would affect the people involved.
Before implementing any AI system, organisations must ask:
- What problem are we trying to solve?
- Is the use of AI necessary and proportionate to the intended purpose?
- What risks could the system create for employees or candidates?
- What information would the AI provide to the decision-maker?
- Can the decision-maker challenge the AI's recommendation?
- Does the decision-maker understand the system's limitations?
- Can an employee or candidate question an AI-influenced decision?
- How will privacy, fairness and transparency be protected?
- Who will be responsible for reviewing and addressing potential harm?
Responsible and ethical AI is not achieved by adopting policies or adding a human at the end of the decision-making process. It requires continuous and thorough attention to the systems being used.
As organisations continue to adopt and integrate AI into their systems, responsible adoption will require more than the latest AI model, it will require a commitment to fairness, human dignity, privacy and trust.
Written by: Syed Tatheer Raza Imam
References:
https://www.ibm.com/think/insights/ai-decision-making-where-do-businesses-draw-the-line
https://www.nist.gov/system/files/documents/2021/08/23/ai-rmf-rfi-0045.pdf?utm_source=chatgpt.com
https://unesdoc.unesco.org/ark:/48223/pf0000381137/PDF/381137eng.pdf.multi
https://www.shrm.org/in/labs/resources/the-evolving-role-of-ai-in-recruitment-and-retention