One year later, he says the gamble paid off. His story has become a real-world test case for what happens when a company moves from human-led operations to an AI-first model almost overnight.
An e-commerce firm that turned into an AI experiment
The company at the centre of this controversy is Dukaan, an Indian start-up that helps small businesses set up online stores. Founder and CEO Suumit Shah had already been leaning heavily on automation, but in the summer of 2023 he crossed a line many executives still fear.
Shah laid off around 90% of his staff and replaced most of the customer support function with AI chatbots. According to his own account, the goal was brutal but simple: cut operating costs sharply and speed up customer service.
In a single restructuring, human support agents were largely swapped for AI systems, turning the company into a live laboratory for automation at scale.
The move instantly triggered criticism from workers, tech commentators and ethics experts. Many questioned whether a start-up that owes its growth to people could so quickly discard them in favour of algorithms.
A year on, the founder calls the results “positive”
Despite the backlash, Shah has spent the last year doubling down on his decision. His assessment of the shift is, in his words, “positive”, especially on the customer-facing side of the business.
Key performance gains he highlights
Based on figures he shared, Dukaan saw notable changes in how support is delivered:
- Average response time dropped from just under two minutes to almost instant replies.
- Time to resolve common issues fell from more than two hours to just a few minutes.
- Support became available at any hour, without the need for shift schedules.
- Costs tied to salaries, training and office space shrank significantly.
Customer queries that once sat in queues now receive automated answers in seconds, with human staff only stepping in for edge cases.
Shah argues that customers welcome this shift because they get faster responses and fewer delays during busy periods. From his perspective, lower costs plus faster support equals a healthier business.
➡️ Expanding AI Career Access Through Skills Based Evaluation
➡️ How to house dairy calves properly?
Critics counter that speed does not tell the full story. They question whether accuracy, empathy and long-term brand loyalty can be measured as easily as response times.
A familiar fear: will AI take all the jobs?
The Dukaan story feeds straight into a broader debate that has been building for years: what happens to work when AI becomes cheap, powerful and always available?
Two opposing camps on AI in the workplace
Reactions tend to fall into two broad groups.
On one side are those who welcome AI as a strong support tool for humans. They see clear benefits in automating:
- Repetitive tasks such as password resets or basic shipping queries.
- Routine data entry and simple report generation.
- Standardised onboarding, FAQs and policy explanations.
For them, AI frees people to focus on specialist work: relationship building, complex problem-solving, strategic planning and creative campaigns.
On the other side are those who see a slow-moving jobs crisis. They point to cases like Dukaan as evidence that once AI is good enough, managers may choose to remove workers entirely rather than simply “assist” them.
For critics, Dukaan is less a story about innovation and more a warning sign of how quickly white-collar roles can disappear.
Where supporters see efficiency, opponents see a future where mid-level employees are hollowed out and only a small group of highly paid specialists remain.
What Dukaan’s experiment actually tells us
Dukaan is far from the only company using chatbots, but few have gone as far, as fast. That makes this case a useful lens, even if it is just one example in a noisy landscape.
Context: customer support is built for automation
Customer support was always likely to be among the first functions hit hard by AI. Many queries are repetitive, rule-based and tied to structured data, which plays to AI’s strengths.
| Task type | AI suitability | Human advantage |
|---|---|---|
| Order tracking and delivery status | High – data is structured and standardised | Low – mostly look-up work |
| Refund eligibility checks | High – rule-based decisions | Medium – edge cases need judgement |
| Angry or distressed customers | Low – emotional nuance required | High – empathy and negotiation skills |
| Fraud or abuse investigations | Medium – pattern detection helps | High – context, legal risk, ethics |
In that light, Dukaan’s numbers on faster responses and faster resolutions are not surprising. The bigger questions lie elsewhere: what happens to the people who held those roles, and what responsibilities do companies carry when they cut staff in favour of machines?
Ethical tension: efficiency against livelihoods
AI does not exist in a vacuum. Decisions like Shah’s sit at the intersection of profit goals, social expectations and individual lives.
Supporters argue that a leaner, more automated company is better positioned to survive in a brutal e-commerce market. If a chatbot can do the same job at a fraction of the price, they say, refusing to use it may be commercially naive.
Detractors point out that workers are often treated as disposable in such transitions. Large-scale layoffs can ripple through local economies, especially when they happen rapidly and without retraining or support.
The friction is clear: investors want margins, customers want instant service, and employees want stable work. AI pushes these demands into direct conflict.
Some policy experts now argue for measures such as reskilling funds, AI taxes, or shared profit schemes to spread the gains from automation more evenly.
What “AI-first” work could look like in practice
Beyond this one company, industries are quietly testing what an “AI-first” structure might be. Rather than starting with people and adding tools, they start with software and add humans only where needed.
Imagine a typical online retail operation using this model:
- An AI agent handles the majority of email, chat and social media queries.
- Only about 5–10% of complex or sensitive cases are escalated to human specialists.
- AI systems produce daily performance reports and recommend pricing tweaks.
- Human staff focus on vendor relationships, product strategy and resolving disputes that carry legal or reputational risk.
In such a scenario, total headcount is far smaller than in a traditional set-up, but the remaining roles are more skilled and better paid. The tension, again, lies in the gap between the number of jobs lost and the number of new roles created.
Key terms and risks worth unpacking
Two phrases often appear in these debates: “chatbot” and “job displacement”. Both hide complexity.
A modern chatbot is not just a scripted FAQ box. It may plug into order databases, payment systems and logistics platforms. That connectivity lets it act rather than just answer. The risk is that bugs, biased training data or misconfigurations can scale into thousands of bad decisions within minutes.
Job displacement, meanwhile, does not always mean instant unemployment. For some workers, AI reshapes tasks rather than replacing them. For others, as in the Dukaan case, roles vanish fast. The speed of change is what creates shock: workers rarely have time to retrain, relocate or negotiate better severance.
Companies considering similar moves can run structured scenarios in advance: what happens if 50% of repetitive work is automated within two years? Which new roles emerge? Which workers can realistically transition? Such exercises help identify training needs before the axe falls.
The Dukaan experiment will not be the last. As generative AI improves and tools become cheaper, more founders will face Shah’s choice: squeeze maximum efficiency from machines, or retain more humans than the spreadsheet demands. The line each business draws will shape not just its balance sheet, but the future shape of work itself.
