Table of Contents
- Key takeaways
- What is conversational AI?
- What are the types of conversational AI?
- 11 conversational AI examples in use today
- 1. Siri AI (Apple): a voice assistant on phones and computers
- 2. Alexa+ (Amazon): a voice assistant that takes actions
- 3. ChatGPT (OpenAI): a general AI assistant
- 4. Erica (Bank of America): a banking assistant
- 5. Klarna AI assistant: customer service for a payments company
- 6. Lyft customer care AI assistant: service for riders and drivers
- 7. Alexa for Shopping (Amazon): a shopping assistant
- 8. Duolingo Max: conversation practice for language learners
- 9. Khanmigo (Khan Academy): an AI tutor
- 10. Erica for Employees (Bank of America): an internal help desk assistant
- 11. Phreesia VoiceAI: a phone agent for medical practices
- What are the most common use cases for conversational AI?
- What does a conversational AI conversation look like?
- What do these conversational AI examples have in common?
- Which type of conversational AI fits your business?
- How can a small business use conversational AI?
- Frequently asked questions
- Basics
- What is an example of conversational AI?
- Is ChatGPT an example of conversational AI?
- Are Siri and Alexa conversational AI?
- Comparisons
- What is the difference between conversational AI and generative AI?
- What is the difference between a chatbot and conversational AI?
- What is the difference between an AI chatbot and an AI agent?
- Use cases
- What are the main use cases of conversational AI?
- What are the applications of conversational AI in customer service?
- Which industries use conversational AI the most?
- Getting started
- How does a business start with conversational AI?
- Does conversational AI replace human agents?
- How much does conversational AI cost?
- Sources
Created by
Do not index
Created time
Oct 25, 2023 11:33 PM
Examples of conversational AI include voice assistants such as Siri AI and Alexa+, general assistants such as ChatGPT, customer service assistants such as Bank of America's Erica and Klarna's AI assistant, shopping assistants such as Alexa for Shopping, and tutors such as Duolingo Max and Khanmigo.
TL;DR
- Conversational AI is software that people talk to in plain language, by text or by voice. This guide lists 11 real examples in five types.
- The most common use cases are customer service, banking, shopping, education, employee help desks, and phone lines.
- Bank of America's Erica has passed 3 billion client interactions. It now averages more than 58 million a month, which is about 1.9 million a day.
- Klarna's AI assistant handled 2.3 million conversations in its first month. Customers resolved errands in under 2 minutes, down from 11.
- A smaller business can use the same idea with a website chatbot trained on its own content, plus a route to a person.
Key takeaways
Type of conversational AI | Real examples | Main use case | How people use it |
Voice assistant | Siri AI, Alexa+ | Everyday tasks and smart home control | Voice, with text as an option |
General AI assistant | ChatGPT | Open questions, writing, and planning | Text or voice |
Customer service assistant | Erica, Klarna AI assistant, Lyft AI assistant | Account, payment, and order questions | Chat inside the company's app |
Shopping or learning assistant | Alexa for Shopping, Duolingo Max, Khanmigo | Product advice, language practice, tutoring | Chat or voice inside the app |
Employee or phone agent | Erica for Employees, Phreesia VoiceAI | IT help and patient phone calls | Internal chat, phone line |
What is conversational AI?
Conversational AI is software that understands what a person types or says and answers in natural language. IBM defines it as "technologies, such as chatbots or virtual agents, that users can talk to".
A conversational AI system does four things in order, according to IBM:
- Takes the input. The person types or speaks through a website, an app, or a phone line.
- Analyzes the input. Natural language understanding works out what the person means. Speech goes through automatic speech recognition first. For a step-by-step look at this stage in a text chatbot, read how an NLP chatbot handles intents, entities, and retrieval.
- Forms a reply. Natural language generation writes the response.
- Improves over time. Machine learning refines the responses as the system handles more conversations.
A chatbot is one kind of conversational AI. Some chatbots follow a fixed script and understand very little language. The difference is covered in chatbot vs conversational AI.
What are the types of conversational AI?
The types of conversational AI are scripted chatbots, AI chatbots, voice assistants, AI agents, and voice agents on phone lines. They differ in how much language they understand and whether they can take actions.
Type | How it works | Good at | Limit | Example in this guide |
Rule-based or scripted chatbot | Follows a script or picks from a library of prepared responses | Fixed, predictable flows | Cannot answer what nobody scripted | Erica draws on a library of more than 700 responses |
AI chatbot | A language model writes each reply, often from the company's own content | Open questions in many phrasings | Needs accurate content to answer from | Klarna AI assistant, Alexa for Shopping |
Voice assistant | Speech recognition plus a language model | Hands-free tasks | Depends on the apps and devices it connects to | Siri AI, Alexa+ |
AI agent | A language model that can also act in other systems | Tasks with several steps, such as bookings and refunds | Errors cost more because the agent acts | Alexa+ books reservations and orders food |
Voice agent on a phone line | An AI agent that answers and places calls | Scheduling, refills, and billing by phone | Needs a handoff for urgent or complex calls | Phreesia VoiceAI |
The line between a chatbot and an agent matters when you buy a tool. The guide to AI agents vs chatbots explains where it falls.
11 conversational AI examples in use today
The 11 conversational AI examples below were chosen on three criteria. Each is live and public, each has a company page or announcement that describes it, and together they cover every type in the table above.
1. Siri AI (Apple): a voice assistant on phones and computers
Siri AI is Apple's assistant on iPhone and iPad. Apple describes it as a conversational AI assistant that people can type to or talk to.
- What it does: answers open questions, finds photos, emails, and notes, and takes actions in apps such as Messages, Music, and Reminders.
- What is new: a dedicated Siri app keeps conversations in one place, so a question started on an iPhone can continue on an iPad.
- Availability: Apple says Siri AI is rolling out in English.
- Type: voice assistant.
2. Alexa+ (Amazon): a voice assistant that takes actions
Alexa+ is Amazon's next-generation assistant, powered by generative AI. Amazon announced it on February 26, 2025 and built it on large language models from Amazon Bedrock. Customers can also use it on the Alexa website and in the Alexa mobile app.
- What it does: controls smart home devices, plays music, makes reservations through OpenTable, and orders food through Grubhub and Uber Eats.
- Scale: Amazon says Alexa+ can coordinate tens of thousands of services and devices.
- Conversation style: customers can pick one of four personality styles, named Brief, Chill, Sweet, and Sassy.
- Type: voice assistant and AI agent.
3. ChatGPT (OpenAI): a general AI assistant
ChatGPT is OpenAI's assistant for everyday questions, writing, images, and code. It works by text, and its voice mode lets people talk with it in real time.
- What it does: explains concepts, drafts messages, searches the web, compares options, and writes and debugs code.
- Why it is a useful example: ChatGPT has no fixed topic. Most business assistants in this list are limited to one company's products and data.
- Type: general AI assistant.
4. Erica (Bank of America): a banking assistant
Erica is the virtual assistant in Bank of America's mobile app. The bank launched it in 2018.
- What it does: answers account questions, shows where a balance is heading over the next 7 days, flags cash back deals, and books appointments with bank staff.
- Published results: nearly 50 million users since launch, more than 3 billion client interactions, and more than 58 million interactions a month.
- The math: 3 billion interactions across 50 million users is about 60 interactions per user. 58 million a month is about 1.9 million a day.
- How it answers: Bank of America says Erica draws on a library of more than 700 responses, and that more than 98% of users find the information they need.
- Type: customer service assistant.
Clients have spent more than 18.7 million hours with Erica. Spread over 3 billion interactions, that is about 22 seconds per interaction.
5. Klarna AI assistant: customer service for a payments company
Klarna's AI assistant answers customer service chats in the Klarna app. Klarna announced it on February 27, 2024, built on OpenAI models.
- What it does: handles refunds, returns, payment questions, and cancellations in more than 35 languages across 23 markets.
- Published results: 2.3 million conversations in the first month, which was two-thirds of Klarna's customer service chats.
- The math: customers resolved errands in under 2 minutes, down from 11. That is at least 9 minutes saved per errand, or about 82% less time.
- Repeat contacts: Klarna reported a 25% drop in repeat inquiries.
- Type: customer service assistant.
Klarna later changed course on staffing. In May 2025, CX Dive reported that Klarna was recruiting people for customer service again and wanted customers to always have the option to speak with a person.
6. Lyft customer care AI assistant: service for riders and drivers
Lyft's customer care AI assistant answers the most common service questions from riders and drivers. Lyft announced on February 6, 2025 that the assistant runs on Anthropic's Claude through Amazon Bedrock.
- What it does: replies to common issues and sends customers to a human specialist when they need more help.
- Published results: Lyft says the assistant cut average customer service resolution time by 87% and resolves thousands of requests a day.
- Type: customer service assistant.
More company examples of this type, including Amtrak, Marriott, and Capital One, are in the list of companies using chatbots for customer service.
7. Alexa for Shopping (Amazon): a shopping assistant
Alexa for Shopping is the assistant inside Amazon's store. It launched under the name Rufus, and Amazon renamed it on May 13, 2026.
- What it does: answers product questions, compares options, and recommends items based on the conversation so far.
- What it is trained on: Amazon's product catalog, customer reviews, community Q&As, and information from across the web.
- Sample questions from Amazon: what to consider when buying running shoes, and how trail and road running shoes differ.
- Type: shopping assistant.
Chatbots for smaller online stores are covered in ecommerce chatbot examples.
8. Duolingo Max: conversation practice for language learners
Duolingo Max is a subscription tier in the Duolingo app with two conversation features, Video Call and Roleplay. Duolingo built them with OpenAI.
- Video Call: the learner holds a spoken conversation with a character named Lily. Lily remembers earlier calls, and the learner gets a transcript afterwards.
- Roleplay: the learner practices scenarios such as ordering coffee in Paris, then gets feedback on accuracy and complexity.
- Human review: Duolingo says its staff write the scenarios and the opening message of each chat.
- Availability: 188 countries, on iOS and Android.
- Type: learning assistant.
9. Khanmigo (Khan Academy): an AI tutor
Khanmigo is Khan Academy's AI tutor for learners and parents. It guides a student toward the answer and does not hand the answer over.
- What it does: tutors on Khan Academy's lessons, coaches writing, and reviews code in JavaScript, HTML, Python, and SQL.
- Voice: learners can speak to Khanmigo, and it reads its replies aloud.
- Price for learners: $4 a month, or $44 a year. The yearly price saves $4 against 12 monthly payments of $48.
- Type: learning assistant.
10. Erica for Employees (Bank of America): an internal help desk assistant
Erica for Employees is the staff version of Bank of America's assistant. It answers employee questions that would otherwise go to the IT service desk.
- Published results: more than 90% of Bank of America employees use it, and it has cut calls to the IT service desk by 50%.
- Type: employee assistant.
Tools for this use case are compared in the guide to IT support chatbots.
11. Phreesia VoiceAI: a phone agent for medical practices
Phreesia VoiceAI is an AI phone agent sold to healthcare organizations. It answers inbound patient calls and places outbound ones.
- What it does: schedules appointments, handles prescription refill requests and billing questions, and routes urgent clinical calls to staff.
- Vendor figures: Phreesia says 80% of calls are handled end to end and that practices see a 30% decrease in patient billing calls. These are the vendor's own numbers.
- Type: voice agent on a phone line.
A healthcare organization in the United States needs a business associate agreement with any vendor that handles patient data. The requirements are in the guide to HIPAA-compliant AI chatbots.
What are the most common use cases for conversational AI?
The most common use cases for conversational AI are customer service, banking, shopping, education, employee help desks, and phone lines in healthcare. The table maps each one to an example above.
Use case | What the AI does | Example | Published figure |
Customer service | Answers common questions and passes the rest to a person | Lyft, Klarna | 87% lower average resolution time at Lyft |
Banking | Answers account questions and sends spending insights | Erica | More than 58 million interactions a month |
Shopping | Compares products and recommends items | Alexa for Shopping | None cited here |
Education | Holds practice conversations and tutors without giving the answer | Duolingo Max, Khanmigo | Duolingo Max is available in 188 countries |
Employee help desk | Answers IT and policy questions for staff | Erica for Employees | 50% fewer calls to the IT service desk |
Healthcare phone lines | Schedules appointments and handles refill and billing calls | Phreesia VoiceAI | 80% of calls handled end to end, per the vendor |
Home and personal tasks | Controls devices, sets reminders, books reservations | Siri AI, Alexa+ | None cited here |
A longer list by industry is in chatbot use cases.
What does a conversational AI conversation look like?
A conversational AI conversation looks like a short exchange in plain language, where the assistant keeps track of what was already said. The sample below is illustrative. It shows a store's website chatbot trained on the store's own help pages.
- Assistant: Hi, this is the store's AI assistant. Ask about orders, shipping, or returns.
- Visitor: hey, do you ship to Canada?
- Assistant: Yes. Standard shipping to Canada takes 6 to 9 business days. The shipping page has the full rates.
- Visitor: and if it doesn't fit?
- Assistant: You can return it within 30 days of delivery. Canadian returns use the prepaid label in the box.
- Visitor: can I talk to someone about a damaged order
- Assistant: Yes. A team member will pick this up. Please leave your email and order number.
Three things in that exchange separate conversational AI from a scripted bot:
- It handles loose phrasing. The visitor wrote "hey" and skipped punctuation, and the assistant still understood.
- It keeps context. The question about fit never mentions Canada or returns, and the assistant connects it to the earlier messages.
- It hands off. The damaged order goes to a person with the contact details attached.
Greetings and small talk work the same way. A good assistant answers a greeting in one line, says what it can help with, and moves to the task. When you write these opening and fallback messages for your own assistant, the chatbot design guide has patterns and example wording for each.
What do these conversational AI examples have in common?
These conversational AI examples have four things in common. Each one has a narrow job, answers from data the company controls, offers a route to a person, and publishes its results.
- A narrow job. Erica answers banking questions. Khanmigo tutors. Only ChatGPT is built for any topic.
- Company data. Alexa for Shopping is trained on Amazon's catalog and reviews. Answering from your own content is called retrieval-augmented generation.
- A route to a person. Lyft sends customers to a human specialist. Klarna keeps live agents available. Erica books appointments with bank staff.
- Measured results. Bank of America, Klarna, and Lyft each published numbers.
The route to a person matters most. A Gartner survey published in September 2026 found that only 27% of customers would try a chatbot again after a negative experience.
Which type of conversational AI fits your business?
The type of conversational AI that fits your business depends on where the answers live, whether the assistant must take actions, and which channel your customers use.
If you need | Pick | Why |
Answers to questions your website and help docs already cover | An AI chatbot trained on your content | It answers many phrasings without a script for each one |
A fixed flow that must run the same way every time | A rule-based chatbot | Every step is predictable and easy to audit |
Refunds, bookings, or account changes done inside the chat | An AI agent connected to your systems | Answering is not enough when the task needs an action |
Help for customers who call | A voice agent on the phone line | A chat widget does not reach callers |
Help for staff with IT and policy questions | An internal assistant trained on company docs | It cuts repeat questions to the service desk |
Complaints, disputes, or anything high-cost | A person, reached through a handoff | Customers rarely give a chatbot a second chance after a bad experience |
How can a small business use conversational AI?
A small business can use conversational AI by adding a chatbot trained on its own website and documents, then routing hard questions to a person. Erica and Alexa for Shopping were built in-house by large companies. A hosted tool gives a smaller team the same pattern without the build. When you are ready to pick one, seven hosted platforms are ranked by human handoff, knowledge sources and price in best conversational AI chatbots.
SiteGPT is one tool of this type. It trains a chatbot on website pages and uploaded files, answers in 95+ languages, and escalates a chat to a person when needed.
- Price: Starter is $59 a month billed monthly, or $39 a month billed yearly ($468 a year). Yearly billing saves $240 a year against 12 monthly payments of $708.
- Quota: Starter includes 1 chatbot, 1,000 pages of content, and 4k GPT-4.1-mini messages a month.
- Trial: 7 days, and a credit card is required. SiteGPT has no free plan.
- Limits: SiteGPT is a text chatbot. It is not a voice assistant or a phone agent. Starter refreshes content manually, and chat integrations such as Slack and Crisp start on Growth at $129 a month billed monthly.
The build steps are in how to create an AI chatbot. Tools that differ on how the handoff works are compared in AI chatbots with human handoff.
Frequently asked questions
Basics
What is an example of conversational AI?
An example of conversational AI is Bank of America's Erica, a virtual assistant that answers banking questions in the bank's mobile app. Other examples are Siri AI, Alexa+, ChatGPT, Klarna's AI assistant, and Duolingo Max.
Is ChatGPT an example of conversational AI?
Yes. ChatGPT is a general conversational AI assistant. People type or speak to it, and it replies in natural language on almost any topic.
Are Siri and Alexa conversational AI?
Yes. Siri AI and Alexa+ are voice assistants, which are one type of conversational AI. Both hold back-and-forth conversations and take actions in apps.
Comparisons
What is the difference between conversational AI and generative AI?
Conversational AI is any system built to hold a conversation. Generative AI is a method that creates new text, images, or audio. Many conversational AI tools now use generative AI to write replies. Erica is conversational AI that answers from a prepared response library.
What is the difference between a chatbot and conversational AI?
A chatbot is a program that chats with users. Conversational AI is the technology that lets a program understand free-form language and keep context. A scripted chatbot with buttons is a chatbot without much conversational AI.
What is the difference between an AI chatbot and an AI agent?
An AI chatbot answers questions. An AI agent also takes actions in other systems, such as booking a table or issuing a refund. Alexa+ is an agent when it orders food through a delivery service.
Use cases
What are the main use cases of conversational AI?
The main use cases of conversational AI are customer service, banking, shopping advice, tutoring and language practice, employee help desks, and phone lines for scheduling and billing.
What are the applications of conversational AI in customer service?
The applications of conversational AI in customer service are answering common questions, tracking orders, handling refunds and returns, collecting contact details, and passing complex issues to a human agent with the conversation attached.
Which industries use conversational AI the most?
The examples in this guide come from banking, payments, ride-hailing, retail, education, healthcare, and consumer electronics. Any industry with a high volume of repeat questions is a fit.
Getting started
How does a business start with conversational AI?
A business starts by listing the questions customers ask most, checking that the website or help docs answer them, and training a chatbot on that content. The next step is a handoff route for questions the chatbot cannot answer.
Does conversational AI replace human agents?
No. Lyft, Klarna, and Bank of America each keep a route to a person. Klarna started recruiting customer service staff again in 2025, a year after launching its assistant.
How much does conversational AI cost?
The cost depends on the type. A learner pays $4 a month for Khanmigo. A website chatbot from SiteGPT starts at $59 a month billed monthly, or $39 a month billed yearly. Custom assistants such as Erica are built in-house and have no public price.
Sources
Every page below was opened and checked on October 6, 2026.
Written by
