CHALLENGES AND REQUIREMENTS OF THE HUMAN RESOURCES PROBLEM
| Challenge | Detailed requirements |
|---|---|
| Resources must be tailored to the specific nature of the work. | Must provide specialized workers; do not assign labeling tasks to programmers. |
| Volume and repeatability | Daily data is voluminous and prone to errors without a strict control process. |
| Phonetic and regional diversity | Workers must be able to understand the local dialect and comprehend the intonation of the Central and Northern regions. |
| High quality, few defects | Write correctly, pronounce clearly, and ensure the recorded version accurately reflects what is heard. |
Candidate requirements:
- Minimum 1 year of telesales experience
- Clear voice, standard pronunciation
- Ability to recognize regional accents (e.g., Quang Ngai, Nghe An, Ha Tinh)
- Listening, comprehension, and writing skills to convert speech into text
SOLUTION
1. Supply of quality personnel
Bellsystem24 Vietnam provides partners with 10 specialized personnel responsible for listening to Voicebot calls—customers and transcribing voice recordings into text (labeling). All personnel are carefully selected and thoroughly trained to meet technical and phonetic requirements.
2. Daily work process
The process is designed to ensure continuity, transparency, and ease of control for partners.
3. Highlights of Bellsystem24 Vietnam's labeling service
Capacity & Service Quality
- Ability to meet staffing requirements and provide suitable personnel
- Ensure performance, accuracy, and output quality as promised (clear SLA, KPI).
Flexibility & Responsiveness
- Personnel, processes, or scenarios can be customized according to specific customer requirements.
- Capable of rapid scaling
Post-Implementation Costs & Services
- Reasonable costs, flexible leasing models (by package, by capacity, by number of seats, etc.)
- Ensure transparency in reporting, monitoring, and evaluating results.
RESULT
- 100% calls received and processed during the day, exceeding the progress commitment
- TAll data has been labeled for model training.
- All record labels are checked and scored. no material errors
- Labeling quality meets AI quality input standards, containing standardized metadata
