Perry Ellis · Yotpo
1.4% — conversion rate
Independently reportedRetail & e-commerce · Customer support · NLP · Scaled
Customer support
AT&T · H2O.ai
Vendor-reportedTelecomIn productionEnterpriseLLM
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 h2o.ai (vendor self-report). Checked against the source page on 2026-09-06. 4 of 4 figures below appear on that page word for word.
The problem
Extracting value and insights from a vast amount of recorded, transcribed, and summarized customer call interactions efficiently and cost-effectively.
What was deployed
- Vendor
- H2O.ai
- Products
- H2O.ai’s Danube 1.8B, H2O LLM Studio
- Models
- GPT-4, Llama, Danube
- Technique
- LLM
- Build or buy
- Bought and customised
- Deployment
- Hybrid
- Data used
- recorded, transcribed, and summarized interactions from 15 million customer calls annually
- Scale
- 15 million customer calls annually
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.
Cost Savings
“AT&T Achieves 90% Cost Savings with Fine-Tuned Small Language Models”
Quoted word for word from the source · round percentage, no baseline given
Latency
“75% Latency improvement”
Quoted word for word from the source · round percentage, no baseline given
91%Accuracy
“The ensemble of small language models achieved 91% accuracy”
Quoted word for word from the source
35% of the previous solutionCostsfrom previous solution
“reduced costs to 35% of the previous solution”
Quoted word for word from the source
Processing time
“a substantial reduction in processing time”
Quoted word for word from the source
Transcript processing capacity
“increased transcript processing capacity”
Quoted word for word from the source
ROI as stated: proven ROI
Sources
- h2o.ai — vendor self-report
Similar deployments
Innovative Dining Group · Revmo AI
1200+ — Reservations booked by AI
Vendor-reportedTravel & hospitality · Customer support · NLP · In production · 2025
Houston Methodist · ActiumHealth
18000000 — staffing capacity created
Vendor-reportedHealthcare · Customer support · LLM · In production
Luxury Bath NJPA · Hatch AI
52% — Overall set rate
Vendor-reportedConstruction & real estate · Customer support · AI agents · Scaled
Radiant Credit Union · interface.ai
60% — Asset growth absorbed without contact center team expansion
Vendor-reportedBanking & finance · Customer support · LLM · Scaled · 2022
Talkmobile · LivePerson
92% — Customer Satisfaction Score (CSAT)
Vendor-reportedTelecom · Customer support · NLP · Scaled · 2014