Free tools Windows power users keep installed
One-click scans. No signup required.
Dukaan founder and CEO Suumit Shah said in July 2023 that the Indian online-store platform had laid off about 90% of its customer-support team after introducing an AI chatbot. That is not the same as replacing 90% of the company’s entire staff. Shah also reported faster responses, shorter resolution times and lower support costs, but the figures were not independently verified. Later stories framed the episode as a one-year assessment without providing a detailed public scorecard showing how service performed over that year.
What happened at Dukaan?
Dukaan is an Indian platform that helps merchants create and operate online stores. On July 10, 2023, Shah publicly said the company had laid off approximately 90% of its customer-support team after deploying an AI chatbot. The distinction matters: coverage of the episode sometimes broadened the claim to “90% of staff,” but the original claim concerned support personnel, not the whole company. The National’s account identifies the affected group as customer support.
Reports identified the assistant as Lina, a Dukaan AI chatbot. Some coverage also connected the episode to Bot9, a chatbot product associated with Shah. Public reporting does not establish how many cases the bot handled end to end, how often customers were transferred to people, or what roles the remaining support staff performed. YourStory’s report describes Dukaan and the Lina/Bot9 connection.
Shah described the decision as difficult but necessary, citing the company’s push toward profitability and challenges in running support. The announcement attracted criticism in part because it presented a large workforce reduction as an efficiency achievement. The available reports do not establish what notice, severance, retraining, reassignment or other transition support affected employees received, so claims about their individual circumstances would go beyond the public record.
#1 Best Overall
What results did Shah report?
The figures below are Shah’s reported before-and-after metrics, as covered by Fortune. They should be read as company claims, not an independently audited performance study.
| Measure | Before, as reported | After, as reported | What the figure describes |
|---|---|---|---|
| First-response time | 1 minute 44 seconds | “Instant” | Time to an initial reply, not necessarily a solution |
| Resolution time | 2 hours 13 minutes | 3 minutes 12 seconds | Reported time to resolution; the measurement method was not published in the cited coverage |
| Customer-support costs | not stated (Fortune, July 12, 2023) | about 85% lower | Shah’s reported cost reduction; the cost categories and calculation were not detailed |
| Support staffing | Human-led support team | About 90% of the customer-support team laid off | Shah’s reported staffing change, not a companywide headcount figure |
On those reported numbers, resolution time fell from 133 minutes to 3.2 minutes—about a 97.6% reduction. That is arithmetic applied to Shah’s figures, not independent evidence that Dukaan’s customers’ issues were resolved that much faster. Business Today also reported the staffing announcement and metrics.
Why faster replies do not prove better support
First-response time measures how quickly a customer receives an initial message. A chatbot can reply immediately and still provide a wrong answer, fail to fix the problem, or make it difficult to reach a person. A resolution-time figure is more useful only if the company defines what counts as resolved and tracks whether customers need to return for help.
The cited coverage does not provide Dukaan’s AI resolution rate, human-escalation rate, repeat-contact or reopen rate, answer accuracy, complaint or refund trends, customer satisfaction, or retention results. Without those measures, the reported speed and cost changes cannot establish that service quality improved or that the bot handled every type of case successfully.
- Response: How long until the first reply?
- Resolution: Was the customer’s underlying issue actually fixed?
- Escalation: How many cases required a person, and how quickly could customers reach one?
- Durability: Did customers reopen cases or contact support again?
- Quality: Were answers accurate, policy-compliant and satisfactory?
Is there a genuine one-year assessment?
Articles published in January and June 2025 presented the episode as a later reflection on the first results. However, the available follow-up coverage largely repeats the 2023 claims rather than publishing a new, independently verifiable year-long dataset. Decatur Metro’s version illustrates that follow-up framing; it does not supply a complete public scorecard for the intervening year.
A useful year-later assessment would show support volume before and after automation, the share of conversations resolved without human help, escalation and repeat-contact rates, customer outcomes, and how many support workers remained. It would also account for the full cost of the system—including development, hosting, monitoring and human oversight—and disclose error patterns and any changes in staffing. Those details are not established in the cited follow-up reporting.
Rank #3
Why the results might not transfer to other companies
There are plausible reasons an AI-first support operation could appear to perform well, but the available reporting does not establish which applied at Dukaan. A platform with a focused product and many repetitive questions may be easier to automate than a business handling varied, account-specific or high-stakes cases. A chatbot can answer many customers at once, while a human team may be constrained by staffing and queue volume. Conversely, an apparent improvement can depend on how “response” and “resolution” were defined or on which cases the bot was allowed to handle.
Cost reduction also needs careful interpretation. If a company dismisses workers, payroll can fall; that alone does not show that AI delivered equal or better customer outcomes. A fair comparison includes technology and integration costs, quality assurance, ongoing knowledge-base maintenance, human escalation, and the effects of service on retention and revenue.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallWhat companies should measure before automating support
For a company considering AI support, Dukaan’s reported figures are not a safe forecast. A controlled pilot should compare AI-assisted support with the existing operation using the same case definitions and issue mix. Track outcomes alongside speed and cost:
- Resolution rate by issue type, including whether the customer’s issue stayed resolved.
- Human-escalation rate, time to escalation and the ease of requesting a person.
- Repeat contacts, reopened cases, complaints, refunds and customer-satisfaction scores.
- Answer accuracy and policy compliance, including errors with financial or account consequences.
- Total cost per successfully resolved case, including engineering, model usage, integrations, monitoring and human review.
- Privacy and security incidents, outages, peak-volume performance and the fallback process.
- Employee outcomes: whether experienced agents are redeployed, retrained or dismissed, and how their case knowledge is retained.
Automation is most defensible for repetitive, low-risk questions when the knowledge base is current, customer data is properly controlled and escalation works. Billing disputes, fraud, account recovery, serious complaints, vulnerable customers and unusual technical cases need a clear route to accountable human review. Poorly governed systems can invent policies, repeat outdated instructions, send customers through deflection loops, or give generic advice when account-specific investigation is required.
What the episode says about AI and layoffs
Dukaan is a notable example of an aggressive AI-enabled restructuring, not proof that AI universally outperforms human support or that companies can safely remove 90% of a support workforce. The original public case rests on Shah’s reported speed, cost and staffing figures; the available coverage does not independently validate them or document customer outcomes over the following year.
The labor question is separate from the service-performance question. Even if automation lowers costs, a responsible assessment should consider how the change was communicated, whether workers had alternatives and whether the company preserved the human expertise needed for exceptions. Public reporting does not establish those details for Dukaan.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
