Designing intelligence with empathy in the age of agentic systems

A few months ago, I sat in a design review meeting where our team was presenting an AI-powered recruitment tool. The tool was supposed to autonomously filter through resumes and select the best possible applicants based on the trained data. The system was built for speed and was technically flawless, scoring candidates with precision, surfacing insights faster than any human could. 

One of our product designers, a woman who had spent a few years in talent acquisition before joining us, raised her hand. ‘This algorithm doesn’t seem correctly optimised,’ she said. ‘Even if it is selecting resumes that look impressive, it is filtering out people who took career breaks, who changed their industries, people whose career paths were not linear.’ She was right.

Historical data show that continuous employment correlates with superior performance. The system consistently downranked candidates with career gaps based on this data. What it missed was context.

Women in India often become caregivers and are forced to take time off work. Some people totally shift their career domains for different reasons. The AI was not discriminating intentionally. It was simply blind to a reality it had never been trained to see.

And in that moment, I realised something I have seen play out repeatedly in two decades of building technology: the systems we design reflect the perspectives of the people who design them. And when those perspectives are narrow, the systems themselves become narrow, no matter how sophisticated the underlying algorithms.

We are living through what many are calling the era of agentic AI. Systems that do not just respond to commands, but act with a degree of autonomy, make decisions, learn from interactions and anticipate needs. These systems are trained on data that reflects the world as it was, not necessarily as it should be. And that is where I believe the perspective of women leaders becomes not just valuable but essential.

Why empathy matters more in Agentic AI

Traditional software follows explicit, auditable rules. Agentic AI, however, doesn't use hard rules. Instead, it examines millions of past examples (training data) to identify patterns. If you are building an AI to hire employees, you feed it 10 years of your company’s past resumes and show it who was selected. The AI "decides" based on what it sees in that history. However, AI doesn’t know about prejudices. If history shows the dominance of males in being selected for leadership roles, it will perpetuate that bias and even amplify this unfairness. 

A biased agentic system might deny a loan, reject a job application or misdiagnose a condition. As they have the power to make decisions that affect real people, the stakes are really higher. So, the need for empathetic design is urgent. 

What women leaders bring to the table: Empathy as a design principle

I am not someone who believes in generalisation. I believe that neither all women think alike, nor that all men lack empathy. But statistically, research shows that diverse teams, particularly those with gender balance, produce more inclusive and human-centred outcomes. A McKinsey study on diversity in innovation found that companies in the top quartile for gender diversity on executive teams were 25% more likely to have above-average profitability. More importantly, they were significantly better at anticipating customer needs and avoiding costly product failures.

Now you may ask why? Because diverse teams hold diverse perspectives, they ask different questions.  More assumptions are challenged, and more missable patterns emerge. And when you are building systems that make autonomous decisions, these points are critical.

One of the most powerful shifts I have observed in how women leaders approach AI design is the emphasis on empathy-driven frameworks. Not treating it as a soft skill, but as a rigorous design principle. This means designing systems that understand context and nuance, not just data and patterns. It means accounting for the unpredictable nature of real human behaviour rather than just the idealised behaviour captured in vast training sets.  

I remember a project for a healthcare client a few years ago. We were designing a patient portal, and the engineering team focused on speed: faster appointment booking and faster test result delivery. These were important metrics. But during user testing, particularly among elderly users and those dealing with chronic conditions, the feedback was harsh. The system felt cold, transactional and inhuman. 

One of our women team members, who had cared for ageing parents herself, suggested a redesign. Instead of optimising only for speed, she introduced what we internally called 'recognition moments': small conversational pauses in which the system acknowledged the user's emotional state before proceeding. 'I understand this must be frustrating. Let me help you find the right person to speak with.' This small shift, barely two seconds of interaction, changed the entire experience. Patient satisfaction scores increased, and retention improved. And critically, vulnerable users stopped abandoning the system mid-conversation.

Three ways women leaders are shaping empathetic AI

So, how does this translate into action? Based on my experience leading digital transformation and observing the teams that build these systems, I see three critical contributions from women leaders.

Championing data equity

The quality of any AI model will be determined by the quality of the data it was trained on. If the data does not accurately represent half the population, it will skew the AI's decision-making process. This was observed in early voice-recognition technology, where systems primarily dominated by male voices could not hear or understand female voices. Whereas women leaders frequently ask questions to ensure the foundation for the agentic system is inclusive and equitable for all.

Designing for emotional context

Contextual understanding is necessary for agentic systems. For example, the phrase “I’m fine” can mean that everything is going smoothly or that it is not; it depends on the tone of voice. However, purely logical systems will always take this statement literally without considering emotional context. In contrast, empathy-driven systems will recognise that "I’m fine" may not necessarily mean the person is indeed fine and will look at supporting evidence, such as whether the person has changed their typing rate, altered their tone of voice, submitted numerous angry complaint letters, etc. Since women leaders have experience working within very complex emotional contexts (both personally and professionally), they are skilled at designing systems that support this type of contextual rationale when designing agentic systems.

Building trust through transparency

We don’t trust a vegetable seller when buying tomatoes. Because we want to know whether it is fresh or rotten. Similarly, while using AI, trust and transparency play a pivotal role. The Agentic system that fails to do so will destroy the transparency relationship, and people will seek other ways rather than use it. Since women are known for trust, in the AI workflow, they are also particularly aligned with keeping transparency. They push for systems that not only make decisions but also communicate why. This is especially important in high-stakes domains like healthcare, finance, and hiring. Users deserve to know why an AI agent recommended one path over another, and they deserve the ability to challenge that decision.

Looking forward

The agentic AI era is not a distant future but a near arrival. The road to failure in AI is paved with good intentions and bad data. But the road to success is lit by diverse perspectives and a commitment to building machines that understand not just our commands, but our real needs. As we stand on the brink of this agentic age, let's make sure we code a little bit of heart into the machine.

(Neha Modgil is Co-founder & COO, TECHVED Consulting)

(Disclaimer: The views and opinions expressed in this article are those of the author and do not necessarily reflect the views of YourStory.)