Categories of services provided by vendors in services using natural language processing

Services provided by vendors

Major categories are bolded, with italicized subcategories following each major category. You can click on a major category to find providers that provide at least one of the subcategories or on a subcategory to find only vendors providing that subcategory.

Customer Service

Customer service options include calling a telephone number, going to a web site and finding a chat option, and interacting with customers through general personal assistants such as Amazon’s Alexa, Google Assistant, or Apple’s Siri. These are most often at least initially automated even when supported by human agents. The use of conversational human language can make such automated services easier-to-use by avoiding a long series of touch-tone menus citing options that never seem to be what you are calling about. The following italicized categories are subcategories of Customer Service:

Call center automation (CCA)

A telephone call to a call center can be automated to meet the caller’s needs directly or to get the caller to an agent with the proper skills. This goal can be achieved with today’s technology, for example, with a prompt such as “Please tell us why you are calling” or “Briefly describe why you are calling.” An automated conversation may ensue if the description is not complete enough to allow action.

Chatbots on web sites (BOT)

Many company web sites feature a chat box where one can ask about a product or get services or information. Some go directly to agents, but, increasingly, an automated system initially starts the conversation, if only to make sure the customer reaches an agent with the appropriate skill set (e.g., standard customer service or technical support). Some chatbots integrate human agents without announcing a specific transfer to an agent. If the automated system is having trouble or can’t address a specific problem, an agent can be sent the chat record and, in effect, continue the conversation. (Many such hybrid conversations end with the customer asking if the respondent is human.)

Company digital assistants reached through the general personal assistants (GPA)

General personal assistants such as Amazon Alexa, Google Assistant, and Apple Siri are in part efficient web portals to information, music, services, etc. (They try to provide a short answer, but often defaulting to a web search if there is a screen available.) They also can control connected devices in a home, such as lights and thermostats. But they are also, like a web browser, a portal to reach specific companies. In this sense, companies will find it increasingly necessary to have a company-specific digital assistant that can be reached through the general digital assistants. If your company provides a service such as home delivery, a user should be able to request delivery through the personal assistants. The personal assistants provide development tools, and there are companies that will develop company apps for them as a service. Amazon calls a digital assistant you add a “skill,” and Google Assistant calls it an “action.” A user can add your digital assistant to the personal assistant repertoire thought an app on a smartphone, or, for example, by saying “Alexa, launch MyCompanySkill.” The personal assistants support more direct access for some company assistants once activated, e.g., “Alexa, ask MyBank for my current balance.” This convenience could become a habit for frequent users of the digital assistants.

Mobile apps (MBL)

A company can connect with a customer directly through a digital assistant reachable through a mobile phone. Customers download your app, and speak to it for customer service, technical support, and in some cases, sales. This is most effective for customers that interact with your company frequently, e.g., for a bank or pizza delivery.

Agent support (AS

Some chatbots are designed to provide agents with information and answers to support their interaction with customers. The customer doesn’t interact with such chatbots directly.

Messaging services

Interacting with customers in natural language through messaging services such as SMS (texting) or Facebook Messenger.

Enterprise Applications

Some bots and assistants are designed for use within companies to support employees. The following are subcategories of Enterprise Applications:

Human Resources support (HR)

Automating some internal human resources activities to ease the burden on in-house support of employee inquiries.

Interfacing with enterprise software (ES)

Making it easier for employees to use enterprise software with natural language interaction.


“Big data” is often an underused asset. Natural language technology can operate on large data sets to classify it into subsets by the language used in a document. If the data is voice, e.g., call center recordings, speech recognition can convert the voice data to text for such analysis. For call centers, this can provide powerful insights into what customers are calling about and flag new issues.

Device and equipment control

Voice control of hardware can simplify complex button-and-screen control systems for hardware devices. As noted, some general digital assistants such as Alexa can control connected devices in the home, but, more generally, a device can have built-in voice control. One advantage of voice control, in addition to its being easier to use in general, is that it can be hands- and eyes-free, allowing control without interrupting activities that require use of hands or eyes. Categories of such applications follow.

Automobile (Auto)

Voice interaction while driving is an increasingly necessary option within automobiles.

In-home device control (DVC)

Home speakers and smartphones can be used to control devices by voice commands. For example, Amazon’s Alexa can command some brands of connected lights, doorbells, thermostats, and TVs.

Industrial equipment (IND)

Some industrial equipment can be controlled by workers using speech when it is difficult to stop what they are doing to use conventional manual controls, e.g., when they are using their hands for other work. Vendors supporting this must provide interfaces to that specialized equipment.

Toys and games (T&G)

Some toys and games benefit from voice interaction. These are typically specialized to the device, and supported indirectly by the core technologies described following embedded in the device. Cloud-based games can, of course, use cloud-based speech technology.

Core technologies

Cloud-based, on-premises, or on-device core technologies (such as speech recognition, speech synthesis, or natural language processing) that allow you to use these technologies as part of an application you are building.

Speech recognition (speech-to-text, SR)

Converts speech-to-text for further processing. The conversion can be fast enough to support a conversation, that is, with a small-enough delay in processing incoming speech from a user that the user experience is similar to talking with a human. The conversion can be much slower if it is essentially transcription, that is, to be read later or archived rather than be part of a conversation. There are other classes of application, such as providing automated subtitles for a video. Some vendors support multiple languages.

Speech synthesis (Text-to-speech, TTS)

Text-to-speech synthesis technology takes text as an input and speaks it in a relatively natural-sounding voice. Typically, a developer can choose a voice and integrate it into an interactive speech system. Some companies will create custom voices for only one company’s use. Some vendors support multiple languages. Some companies provide TTS voices to be used by only one company. Sometimes called “cloning,” custom voices are created by taking a small sample of a voice a company wants its TTS software to sound like, and creating a TTS version that speaks any text in that voice. Some providers have “off-the-shelf” custom voices, that is, they have them pre-built so that you can listen to them and buy exclusive use of one you like.

Natural Language Processing

Automated interpretation of text to understand its intent and relevant content, natural language “understanding.” (The text may have been created by speech-to-text processing, and some companies indicate they analyze speech directly for its content.) The purpose of NLP is to allow taking action on content expressed in natural human language. Some vendors support multiple languages.

Speaker recognition (SKR)

Detecting the identity of a speaker from previously registered speech characteristics, this technology can be used for applications such as displaying who is speaking a virtual meeting or for biometric identification for security.

Speech enhancement

Improving the quality of a speech signal by, for example, reducing background noise.

Machine learning (ML)

Machine learning is a basic AI technique that can be used in other core technologies, including natural language processing and speech recognition. Some companies offer it as an independent service.

Content (CNT)

Content is information and entertainment services that could be used to provide audio or text responses that service a user request.

Voice Recording

High-quality voice recording for prompts or for creating custom Text-to-Speech voices.

Natural Language Generation (NLG)

NLG generates readable text from data, providing a natural-language interpretation of a table of data or data from a database.


The “scope” of a company’s service or product is the degree to which it can develop a full digital assistant or chatbot without a narrow specialty (e.g., only skills for Alexa) or a pure technology (such as speech recognition). The different cases will be designed “Full” or “Specialized” (SPC) in tables. This is an inexact label, and can change as services evolve.




When a vendor builds a bot or digital assistant for a company or provides other services, such as analytics, the solution may be executed in the cloud. Some companies want to host at least some customer information on their own premises or execute the entire service on-premise or even on a specific device.

Cloud-based (CLD)

At least one version of the vendor product is fully cloud-based.

On-Premises (PRM)

At least one version of the vendor product is a full customer-premise-based option.

Hybrid (HYB)

Some vendors have hybrid options, where much of the technology is cloud-based, but specific customer or other sensitive data can be limited to company premises.


Some technologies can run on “edge” devices such as smartphones, rather than in the cloud or on company servers. A special case is “wake-up” words that can turn on listening in a device that otherwise uses a cloud-based services; the wake-up work, such as “OK, Google,” avoids all speech that the device hears being sent to the cloud to see if it is addressed to the digital assistant, an obvious privacy issue that also makes the support of the assistant much less expensive.


Some bots are designed to specifically support customer service agents. They thus must work within customer service platforms.


Because language coverage changes often and can cover tens of languages, it is necessary to go to the vendor web site or contact the vendor. Almost all vendors support English.

Vendors providing a category of service

Click on a category of service to find vendors supporting that category of service. (This feature is currently under development.)