Unilever · Google Cloud
15-35% increase — retailer sales
Independently reportedOther · Other · Classical ML · Scaled
Other
Thanaka Bangladesh · Karigor AI Labs
Vendor-reportedRetail & e-commerceBangladeshIn productionMid-marketNLPLive 2026
Vendor-reported. The customer is named and the numbers are quoted from the source page, but the account comes from the vendor. No independent confirmation.
Reported by karigor.ai (vendor self-report). Checked against the source page on 2026-09-05. 5 of 10 figures below appear on that page word for word; the rest are marked.
The problem
Sales capacity was limited by human staff, leading to lost leads due to slow response times, especially during off-hours.
What was deployed
- Vendor
- Karigor AI Labs
- Technique
- NLP
- Build or buy
- Bought off the shelf
- Deployment
- SaaS
- Data used
- Product catalog with prices, descriptions, benefits, usage instructions, SKUs; Bangla conversation style, formal register, brand voice.
- Scale
- 1,344 conversations in 13 days
What changed
Each row is quoted from the source. Figures we could not find on the page in those words are marked — they are kept, not deleted, so you can judge them.
1,344Conversations handled13-day window
“Over a 13-day window it handled 1,344 real customer conversations in Bangla”
Quoted word for word from the source
98.6%Conversations handled end-to-end by agent
“98.6% of them were handled end-to-end by the agent (as measured, April 2026)”
Quoted word for word from the source
Under 10 secondsResponse time
“Agent responds in under 10 seconds with relevant product information”
Quoted word for word from the source
Conversations lost to slow responsefrom ~40%
“Conversations lost to slow response ~40%”
Not found on the source page in these words
4.2%Human handoff needed
“Human handoff needed 4.2% (2 of 48)”
Not found on the source page in these words
100,000+Estimated revenue recoveredmonth
“Estimated revenue recovered ৳100,000+/month (previously lost to slow response)”
Not found on the source page in these words
~40%Ad spend efficiency gain
“Ad spend efficiency gain ~40% of leads no longer wasted”
Not found on the source page in these words
14Off-hours coverage addedday
“Off-hours coverage added 14 additional hours/day”
Not found on the source page in these words
95%Product match accuracyfrom ~90% (human staff)
“Product match accuracy | 95% | ~90% (humans misidentify too) | —”
Quoted word for word from the source
48+ (unlimited)Conversations handled per dayfrom ~15-20 (business hours only, human staff)
“Conversations per day | 48+ (unlimited) | 15-20 | Varies”
Quoted word for word from the source
ROI as stated: Payback period Under 1 week
“We were paying for ads that brought customers to our door, then losing them because the door was closed.”
Difficulties and limits
- Sales capacity was limited to the number of conversations staff could handle simultaneously.
- Peak hours were evenings and late night when staff was off.
- Average response time during off-hours was 2-8 hours, leading to ~40% conversations lost.
- Previous attempts with night-shift staff were too expensive.
- Facebook's built-in auto-replies lowered customer trust.
- Rule-based chatbots failed on Bangla misspellings and couldn't handle complex interactions.
- The agent hands a conversation to a person only where its rules said to: a complaint, a bulk order, a price negotiation, or when the customer asked.
Sources
- karigor.ai — vendor self-report, 2026-03-28
Similar deployments
ANZ Bank · NVIDIA
0.82 — Gini coefficient for assessing risk
Independently reportedBanking & finance · Other · Classical ML · Pilot
BHP Billiton
zero — accidents due to driver drowsiness
Independently reportedOther · Other · Classical ML · In production · 2022
DSST: College View · Playlab, Class Companion, Manta AI
Vendor-reportedEducation · Other · NLP · In production · 2024
Cubo Ai · Google Cloud
more than 10X — user growth supported with same IT workforce
Vendor-reportedOther · Other · Computer vision · Scaled · 2019
Precision HVAC & Cooling · AutoScale My Biz
15+ — High-Value After-Hours Leads Captured
Vendor-reportedOther · Other · LLM · In production