In 2023, an Indian e‑commerce boss sacked almost his entire customer support team and handed their work to chatbots.
A year on, he says the gamble has paid off. The numbers look clean and efficient, yet the human cost and long‑term risks remain anything but clear.
An e-commerce startup that went all-in on automation
The company at the centre of this experiment is Dukaan, an Indian platform that helps small businesses sell online. Its co‑founder and CEO, Suumit Shah, has become one of the loudest advocates of aggressive AI adoption in the retail tech sector.
In the summer of 2023, Shah decided to cut around 90% of the firm’s support staff and replace them with AI‑powered chatbots. The move landed him in the global spotlight, not for a funding round or a flashy acquisition, but for a radical workforce shake‑up that many saw as a warning sign for white‑collar work.
Shah framed the decision as a hard but rational step for a bootstrapped company under pressure to reach profitability. Wages, training and churn in customer service were eating into margins. Generative AI tools were getting cheaper and better. The temptation to automate was strong.
Within months, human agents who once handled complaints, refunds and basic queries were gone, replaced almost entirely by software running 24 hours a day.
The change was not gradual. It was swift, public and unapologetic, sparking anger from workers and fascination from tech investors who saw Dukaan as a live test case.
A “positive” report card from the CEO
Now, roughly a year after the layoffs, Shah says the bet on AI has worked from a business standpoint. The first internal assessment, which he has discussed in interviews and social media posts, highlights three headline gains.
Faster replies, shorter queues
Before the switch, customers typically waited close to two minutes for a support agent to answer a query. That lag is familiar to anyone stuck watching a spinning wheel on a live chat window.
With chatbots handling the bulk of incoming messages, Dukaan now claims responses are “almost instantaneous”. The bot engages as soon as the user types, pulls up account data and guides them through menus or suggested answers.
➡️ A study suggests cats can develop a dementia-like condition similar to Alzheimer’s
➡️ Why your mother’s cells stay inside you for life
➡️ Six types of gifted brains, explained by American experts: which one sounds like you?
➡️ 7 phrases older than 65 use that sound totally out of touch to young people
➡️ A “flesh-eating” bacteria found in some oysters kills two people
➡️ Forget olive oil: this alternative is far healthier and more affordable
According to Shah, average resolution time for common problems has dropped from more than two hours to just a few minutes.
For customers who just want a refund confirmed or an order tracked, that speed can feel like an upgrade. There is no small talk, no shift change, no “please hold while I check with my supervisor”.
Lower costs and leaner operations
On the financial side, the appeal is obvious. Fewer employees mean lower payroll costs, less office space and reduced spending on recruitment and training.
- About 90% of support roles were cut in 2023.
- Chatbots now handle the majority of routine questions.
- Human staff mainly manage exceptions and system oversight.
Shah argues that the savings have helped Dukaan sharpen its focus on product development and sales, instead of running a large operations team. He presents this as a necessary step in a fiercely competitive e‑commerce market with thin margins and impatient investors.
Ethical questions that refuse to go away
For critics, the figures do not tell the full story. Around the same time that response times improved, dozens of workers suddenly found themselves out of a job in a sector already under pressure.
Commentators in India and abroad have questioned the ethics of using AI not as a complement to human workers but as their direct replacement. The Dukaan case is often cited in union meetings and policy debates as an early example of what mass white‑collar automation could look like.
The tension sits between two narratives: AI as a tool that frees humans from drudgery, and AI as a blunt instrument for cutting payroll.
Both narratives contain some truth, and Dukaan has become a convenient symbol for the second.
Supporters vs sceptics
The public conversation tends to split into two broad camps.
| Viewpoint | Main argument |
|---|---|
| Pro‑AI adoption | Automation removes repetitive work, cuts costs and lets people focus on higher‑value tasks. |
| AI sceptics | Large‑scale replacement of people by software threatens jobs and weakens worker bargaining power. |
Pro‑AI voices say that resisting these tools will only push companies toward competitors willing to use them. They point to the history of technological change: from factory machines to spreadsheets, new tools have always disrupted older roles.
Sceptics counter that generative AI moves faster and cuts deeper than previous waves of automation, especially in services once considered safe. They argue that without strong safety nets and retraining, the losers in this shift will be very visible, while the winners stay concentrated in a small circle of founders and shareholders.
What AI can and cannot do in customer service
The Dukaan case also exposes the limits of today’s chatbots. While they excel at routine questions, they can stumble in messy, emotional or unusual situations.
Customers dealing with lost packages, fraud or complex billing issues often want empathy and negotiation, not canned responses. Even the smartest model can misinterpret tone, escalate frustration or produce confident but wrong answers when pushed outside its training data.
Behind the impressive metrics, a small human team still needs to monitor the system, handle edge cases and intervene when the bot gets stuck.
That hybrid model — AI for volume, humans for nuance — is where many analysts think support operations will settle, at least for the next few years.
What this means for workers and businesses
For employees in sectors like e‑commerce, banking and travel, Dukaan offers a snapshot of possible futures. Routine support roles, ticket triage and basic back‑office tasks sit at the top of the automation risk list.
At the same time, new roles are emerging around these tools: prompt designers, AI trainers, workflow architects and specialists who review bot decisions for compliance or bias. These jobs require different skills, from data literacy to soft skills for managing escalations.
For businesses, the trade‑offs run beyond payroll. Reputational risk rises when a brand is seen as ruthless with staff. Mistakes by chatbots — such as mishandling complaints or giving inaccurate advice — can travel fast on social media and damage trust.
Key risks for firms racing into AI-only support
- Loss of human insight into customer pain points.
- Greater exposure to technical outages or model errors.
- Regulatory scrutiny if automation leads to unfair treatment.
- Lower staff loyalty if workers feel replaceable at short notice.
Reading the Dukaan experiment as a warning sign
The Dukaan story is already influencing boardroom conversations far beyond India. Leaders see the efficiencies and feel the pressure to do something similar. Workers see the layoffs and wonder if their job is next in line.
Two terms pop up constantly in these debates. “Generative AI” refers to systems that create new content — text, images or code — rather than just classifying data. “Chatbot” describes the conversational interface that sits on top, giving the illusion of a human‑like assistant while the real work happens in large models behind the scenes.
Both are advancing quickly, and both are still fallible. That gap — between promise and reality, between cost savings and social impact — is exactly where the Dukaan case sits. Companies watching from the sidelines will be weighing not only the numbers, but the kind of workplace and customer relationships they want five years from now.