One year on, his experiment in radical automation is starting to reshape how tech leaders think about people, profits and the limits of artificial intelligence at work.
An Indian startup that bet almost everything on AI
The decision came from Suumit Shah, founder and CEO of Dukaan, a Bangalore-based platform that helps small merchants run online shops. Facing tight margins and fierce competition, Shah chose a brutal route: he dismissed around 90% of his staff, mainly in customer support, and replaced them with AI-powered chatbots.
At the time, the move sparked outrage on social media and among labour advocates. Many saw it as a warning shot of what could happen across the global customer service industry. Others framed it as a cold but logical step for a lean, venture-backed startup under pressure to grow.
Shah’s stated goal was simple: cut costs sharply while making support faster and more efficient for customers.
Customer support teams are often the first to face automation. They handle repetitive queries, rely on scripted answers and operate under intense pressure to be both cheap and available around the clock. For an AI system, that combination is attractive.
A “positive” first-year report, according to the CEO
One year after the layoffs, Shah has now shared his first proper assessment. From his perspective, the gamble worked.
According to figures he highlighted, response times to customer questions dropped from nearly two minutes on average to almost instant replies. Complaints and order issues that once took more than two hours to resolve are now often wrapped up within minutes.
The company reports near-instant answers and much faster problem resolution, at a fraction of the former staffing cost.
For an online seller, that kind of speed can translate into fewer cancelled orders, fewer angry buyers and potentially higher repeat business. Dukaan claims its users appreciate the faster service, though independent data on customer satisfaction has not been widely published.
On the financial side, cutting salaries and office costs dramatically improves the short-term picture for any startup. Support teams are expensive: they need training, supervision, HR support and physical or virtual infrastructure. A chatbot, in theory, needs only cloud computing, updates and occasional fine-tuning.
What exactly changed inside the company?
The transformation at Dukaan was not just a simple swap of people for software. It required redesigning workflows around the AI systems and rethinking what human employees remain for.
From human-first to AI-first support
Before the shift, a typical customer interaction at Dukaan went through several steps: a user raised a ticket, a human agent responded, escalated if required, and sometimes involved managers or technical staff. After the overhaul, the chatbot became the front line and, in most cases, the only line.
- The bot greets customers and identifies the problem.
- It pulls information from order databases and help files.
- It suggests fixes or processes refunds and changes when allowed.
- Only complex or unusual issues are escalated internally.
Shah has said that the vast majority of contacts can be solved without human involvement. That claim, if true, suggests that a large chunk of previous support work was highly repetitive and rule-based.
How the metrics look
Although Dukaan has not released a full set of audited figures, the company emphasises three main metrics:
| Metric | Before AI rollout | After AI rollout |
|---|---|---|
| Average first response time | Just under 2 minutes | Almost instant |
| Average resolution time | Over 2 hours | Several minutes |
| Support staffing level | 100% of original team | ~10% of original team retained |
These numbers fit a broader pattern seen in early AI deployments: dramatic speed gains, clear cost reductions, but still limited transparency on long-term customer loyalty and brand perception.
A polarising case study in AI replacing people
The Dukaan story crystallises a question hanging over many industries: when does AI assist humans, and when does it simply replace them?
Supporters of AI in business argue that systems like chatbots remove boring, repetitive duties. They claim this frees human workers to take on creative, strategic or relationship-based roles. They also point to lower prices for consumers and better service availability, especially outside normal office hours.
For AI enthusiasts, Dukaan is proof that software can handle high-volume customer work faster, cheaper and at scale.
Critics see something very different. They worry that if one CEO can remove 90% of a team and publicly celebrate the result, other firms will follow. That could put tens of thousands of support jobs at risk across markets where call centres and help desks are major employers.
Ethical questions also surfaced when Shah initially announced the layoffs. Many observers asked whether the company had done enough to retrain or redeploy affected staff. Others questioned the social cost of celebrating such a deep cut simply as a “productivity win”.
What this means for workers and businesses
Dukaan’s experience will be closely watched by startups and big corporations alike. Customer service, content moderation and basic back-office work are already being targeted by generative AI tools.
For workers, especially in countries like India or the Philippines that host huge support centres, the case highlights the need for new skills. Employees who can supervise AI systems, handle complex customer situations or design workflows have a better chance of staying in demand.
For companies, the lesson is more nuanced. Replacing people with software can slash costs, but it also changes the brand’s relationship with its users. A chatbot that is fast yet rigid can frustrate customers who have unusual problems, emotional complaints or sensitive issues.
When every competitor offers instant AI replies, companies may compete again on something deeply human: empathy and trust.
Key terms and concepts behind the Dukaan experiment
Several jargon-heavy ideas sit behind Shah’s headline-grabbing move. A few are worth unpacking.
Chatbot vs. human agent
A chatbot is a computer program that simulates conversation through text or voice. Modern chatbots, powered by large language models, generate responses rather than picking from a fixed script. A human agent, by contrast, relies on training and judgement, and can flex rules in ways AI still struggles to match.
When businesses weigh one against the other, they usually balance:
- Cost per interaction
- Speed of response
- Accuracy of information
- Customer satisfaction and loyalty
- Regulatory and ethical risks
Automation risk scenarios
If a mid-sized e‑commerce firm follows Dukaan’s path and replaces most of its support staff with AI, several scenarios can play out:
- Short-term boost: Costs drop quickly, profit margins improve and investors react positively.
- Mixed customer reaction: Routine queries become smoother, but edge cases and emotionally charged complaints cause friction.
- Reputational shift: The brand gains a tech-forward image, yet may be seen as less human or less caring.
- Policy pushback: If mass layoffs become common, regulators and unions could push for new rules on AI-driven restructuring.
Risks, benefits and what could come next
The Dukaan case illustrates both clear benefits and meaningful risks of aggressive AI adoption.
On the benefit side, companies can expand support hours, handle heavy traffic during sales and cut operational spending. For fast-moving startups, this may be the difference between surviving and shutting down.
On the risk side, over-reliance on AI raises concerns around bias, data privacy and error handling. A misconfigured bot can make the same mistake thousands of times before anyone notices. A disgruntled customer who never reaches a human may walk away for good, then share that experience widely on social platforms.
Some analysts expect a hybrid model to become the norm. Under that approach, AI would manage simple queries, while smaller, highly trained human teams tackle sensitive, high-value or complex cases. Dukaan’s extreme approach pushes much further, but its results will feed into boardroom debates across the tech and retail sectors.
For now, Suumit Shah maintains that his decision was justified by the numbers. For laid-off staff, and for millions of workers in similar roles worldwide, the real outcome will be measured less in response times than in the availability of new, decent jobs in an economy that is learning to live with — and sometimes replace — human labour using AI.
