All You Need to Know About AI Call Center Industry Jargon
-
This article explains the 'industry jargon' of AI call centers. Those in the field can gain insights, while others can enjoy the read.
During the pandemic, many businesses couldn't resume operations and had to rely heavily on phone calls to keep their businesses running. Due to the management challenges of manual telemarketing, AI telemarketing robots became extremely popular. These robots have now evolved to mimic human voices so well that their response speed and conversational abilities are nearly indistinguishable from humans. If you receive a very polite, well-articulated sales call, there's a high chance it's a robot.
I’ve worked on AI call center products myself and even set up a voice assistant for my phone to handle calls when I’m busy. Recently, many people have asked me about AI call centers, so today, let’s dive into the 'industry jargon' of AI call centers. Those in the field can gain insights, while others can enjoy the read.
*AI Robots vs. Human Agents (Source: DataPoint Marketing Official Website)
Before diving into the jargon, let’s first outline the workflow of an AI call center system to give everyone a comprehensive understanding.
Workflow of an AI Call Center
The diagram above shows the workflow of an AI call center. Simply put, the first step is to configure the script—telling the robot how to respond to different scenarios. This varies significantly across companies and industries, so each client typically gets a customized script.
After configuring the script, the next step is to set up phone lines for the account to ensure calls can be made smoothly.
Once the lines are set up, the next step is to create tasks—essentially importing phone numbers. Due to limited resources, it’s impractical to dial all numbers at once, so tasks are assigned to robots to complete one by one. Notably, if multiple lines are available, it’s advisable to include a line selection feature when creating tasks. Many businesses operate in multiple cities, and people are more likely to answer calls from their local area, so this feature helps improve connection rates.
After completing the calls, the robot evaluates the intent of each call—essentially the feedback post-call. Typically, the system includes statistical and query functions to analyze the results of each call.
Now that we’ve covered the workflow, before introducing the jargon, I’d like to add one more point. Despite their advancements, telemarketing robots are still quite immature. Many business owners have misconceptions about them. In my view, the core role of telemarketing robots is to improve the efficiency of sales teams by performing initial screenings—identifying potential clients for human follow-up—rather than closing deals on their own, as some might believe.
Alright, with that out of the way, let’s get to the jargon. I’ve divided it into two parts: industry jargon and technical jargon. Industry jargon refers to professional terms related to the business, while technical jargon covers the technical aspects. To prevent you from being misled in the future, I’ll explain both.
Script
As mentioned earlier, this is similar to sales scripts—telling the robot how to respond to questions. The only difference is that these scripts need to be configured into the system, and the communication logic must be clearly structured. Scripts vary by company and industry, so many robot providers customize scripts for each client.
Human-like Voice
Many robot vendors claim to offer 'human-like voice,' but this doesn’t mean an actual person is making the calls. A more accurate term would be 'human voice dubbing.' Essentially, a real voice actor records all responses, which are stored in the cloud. When the robot encounters a relevant question, it plays the pre-recorded audio, making the caller feel like they’re talking to a real person.
Phone Lines
These are the channels through which robots communicate with clients—essentially phone numbers. Only by establishing a connection via a number can a conversation occur. This has always been one of the most challenging aspects of AI call centers, not because the technology is complex, but because regulations against spam calls are strict, making it hard to secure stable lines. Some companies use gateways to address this issue. The most impressive setup I’ve seen involved a 128-port gateway dialing simultaneously—though the cost is quite high.
128-Port Gateway (Prices typically exceed ¥10,000)
Card Registration
This isn’t about membership cards but phone SIM cards. Due to strict regulations requiring real-name registration, individuals can only register a limited number of SIM cards. This leads to situations where companies buy telemarketing robots but quickly run out of usable SIM cards. Some companies now offer robots that don’t require clients to register SIM cards, effectively solving this problem.
Connection Rate
This refers to the ratio of answered calls to total calls made. It depends on various factors, including number quality and location. Even with robots, there’s a failure rate—yes, you heard that right. At one point, many robots on the market had success rates below 20%.
Caller ID Display
This is simply the number displayed on the recipient’s phone. Most people prefer answering calls from mobile numbers rather than landlines, so mobile caller ID is more common, though a few clients may request landline displays.
Mobile or Landline—Which Would You Answer?
ASR (Automatic Speech Recognition)
This is the technology that converts speech to text, functioning like human ears. The performance of ASR depends on four main factors.
NLP (Natural Language Processing)
This refers to natural language understanding, similar to how the human brain processes information. After the 'ears' (ASR) translate speech into understandable text, the 'brain' (NLP) interprets the meaning—whether it’s agreement or a question. NLP is one of the most challenging areas in AI, and its techniques for parsing natural language are quite fascinating.
TTS (Text-to-Speech)
This is speech synthesis technology. Voices like Siri or Xiaomi’s Xiao Ai are generated using TTS, though they still sound somewhat unnatural. In telemarketing robots, TTS is sometimes necessary—it’s impractical for voice actors to record every possible response, especially when confirming account or phone numbers.
Siri—A Classic Example of TTS Technology
Noise Reduction and Voice Enhancement
As mentioned earlier, one factor affecting speech recognition is audio quality. Noisy environments significantly reduce accuracy. However, many noises operate at different frequencies than human speech, allowing technologies to filter them out. Voice enhancement techniques further refine the audio for ASR processing.
Empty Number Detection
To improve dialing efficiency, some providers pre-screen numbers to filter out inactive ones, though current success rates for this technology aren’t very high.
That’s all for now!