‘Alexa+ is more conversational’: Amazon execs on bringing AI-powered assistant to India | Technology News

However, voice-based interactions are different from text-based chatbots. “Imagine if every time you had to wait to read through the entire answer. It would be very slow. The way the voice experience works is actually quite different from the way a chatbot can answer with graphs, images, numbers, long answers and bullet points.”

‘Wanted to be authentically Indian’

Announced last year and rolled out in the US earlier this year, Amazon, the world’s largest company by revenue and perhaps the biggest e-commerce platform, is bringing Alexa+, its next-generation Alexa voice assistant powered by artificial intelligence, to India.

The Generative AI-powered assistant promises more conversational responses, a better understanding of complex requests much like a Chatbot, and has a natural-sounding voice over Alexa’s robotic tones.

“We wanted to be authentically Indian and be able to understand the culture, the sense of humour, the history, the values and our customers here. Alexa+ is not a chatbot or a single large English model. It has to do much more. It has to take actions in the real world. It has to be able not just to play music, but now also to book a restaurant,” Butti went into detail, explaining the effort it took to localise Alexa+ for Indian tastes.

Available as part of the Early Access programme, Alexa+ is offered in English, Hindi, and Hinglish as an opt-in feature on screen-equipped Echo Show devices. After the Early Access period ends, Alexa+ will be free for Prime members, while non-Prime users will pay Rs 2,000 per month for a standalone subscription.

Michele Butti, Vice President, Alexa International (left), and Teena Sidana, Country Manager, Alexa, Amazon India (right), were in New Delhi for the launch of Alexa+ in India. (Image credit: Anuj Bhatia/The Indian Express)

Fine-tuned for India

While users can do many of the things they could already do with the original Alexa, the revamped voice assistant feels different. It is built with generative artificial intelligence and feels much more modern, akin to ChatGPT or Gemini.

“It’s much more conversational. Being based on generative AI, it can talk about anything, not just the deterministic technology we had before, where we had to script every question, the right answer, and what we did. Because people don’t converse by using a wake word or a name for every single term, you can have a longer conversation without using the Alexa wake word every time. The way it works, though, is designed to make sure that privacy is protected and customers stay in control of what Alexa listens to and what it doesn’t,” Butti explains.

This is more of a move to compete with chatbots like OpenAI’s ChatGPT and make Alexa a more helpful assistant for users’ day-to-day needs.

“The technology we use is our own Automatic Speech Recognition (ASR) that is fine-tuned for Indian homes today,” Teena Sidana, Country Manager, Alexa, Amazon India, says.

“In a chatbot, where you are just typing or speaking, you don’t need the ability to deal with a noisy home in the background. It’s one person, one device. You type your answer, you get a longish response, and you’re okay with it. But with a voice-based interaction, everybody’s in the background, there’s a pressure cooker, the TV is on, and there’s a child talking. The microphones have to be fine-tuned to listen only to that person, that wake word and that command, and adapt to the household,” Sidana says.

“There are colloquialisms and the way we mix languages when we talk. Nobody speaks in proper English or proper Hindi all the time. Everything is mixed-language and code-mixed in India. So, all of these nuances that are unique to India also had to be taken into account when it comes to voice — how it handles code-mixing, how it talks and how it reacts,” she adds.

“You need to have that voice. You have to have the right emotions. If I am sad, she needs to recognise that and respond: ‘Oh, you are sad, let me give you some nice music to cheer you up.’ If I am happy, it’s like, ‘It’s a great day for you. Let me help you with something upbeat. Do you want to celebrate today and order your favourite snacks?’ So that is, I think, the true personality or use of Alexa, where she can adapt to your choices, your preferences, your household and the local, authentic Indian context around you.”

“We had to make sure that when we brought Alexa+ to Indian customers, it was already connected with all the services and content and was able to keep working with multiple smart home providers, including the majority being actually Indian manufacturers and providers,” Butti adds.

‘Understands customer’s intent’

Along with a visual experience, Alexa also offers a much improved audio experience.

“We have developed far-field technology that helps Alexa work not just when you hold the microphone in front of your face, but also when the device is far away from you. It can still isolate what you are asking because of the unique combination of hardware and multiple microphones that can isolate the source of the voice from all the background noise,” Butti says.

Butti said Alexa+ is being designed to understand users’ intent, and that the experience with the new Alexa will be different from the original Alexa.

“When we started at the very beginning, without customising the Alexa experience, we noticed that customers felt it was overly verbose. Now, we have trained Alexa to better recognise the customer’s intent. When you ask Alexa, ‘Turn on the light,’ you are not there to chat. Sometimes, it doesn’t even answer; it just executes the command, and that’s it,” he says.

“But when you ask, ‘What’s new around the world?’ then you are in a different state. Based on that, she’ll tailor her response and give you a quick answer. If you ask, ‘What was the score of my favourite team yesterday?’ — boom, you get the answer,’” he continues.

“If you say, ‘I had a really heavy day, and my boss made my day miserable,’ that’s a very different intent. The beauty of this technology is that it’s so much smarter that it can understand the customer’s state and respond accordingly.”

“The ability of an LLM or generative AI is that it can adapt to the way you want the responses to be,” Sidana says.

The revamped Alexa uses memory to maintain conversational context and personalise interactions across diverse household needs. It remembers user preferences and adapts its responses, letting users interrupt or change requests naturally without repeating information. It can also store specific user instructions, such as dietary preferences, for better personalisation over time.

Alexa+ works with Amazon’s new Echo devices, Fire TV, and the Alexa app. (Image credit: Amazon)

‘Much easier way to interact with AI’

Butti agrees that while many competitors overemphasised AI in messaging, AI should remain in the background.

“I think you don’t need to understand or know anything about AI. It’s so intuitive. Sure, it’s entirely based on LLMs and Gen AI, but the point is that it has this super intelligence that allows you to go very deep into science or talk about sports,” he says. “With Alexa+, it’s a much easier way to interact with AI and get the benefits of this new technology than using a chatbot, where you still have to download a dedicated app, learn what works and what doesn’t, and figure out which model you need to use,” he adds.

Sidana emphasises that Alexa+ has always been designed for home use. “There are going to be multigenerational households. There will be somebody who listens to their morning bhajan on it and there will be a kid who asks for help with science, their schedule or their calendar. It is designed for anybody and everybody to interact naturally with the device and ask for whatever they need or whatever their preferences might be, without having to learn something entirely new. It’s not a coding device. It’s not just a science-learning device. It’s not a PhD device or an app,” she says.

With Alexa+, Amazon is also dipping its toes into agentic AI, allowing the service to perform real-world tasks on users’ behalf, such as ordering food, booking restaurants, and buying gifts. The company has partnered with local services such as Swiggy and Zomato for food delivery, EasyDiner for table bookings, and Chef Sanjeev Kapoor for recipes, as well as news providers such as PTI and India Today for region-specific news updates.

While the smart home as a concept is still in its nascent stages in India, Sidana said interactions with Alexa through smart home devices have grown, thanks to the expansion of affordable smart devices across Indian households. More than 1.1 billion smart home interactions took place over the past year across devices such as bulbs, plugs, switches and air conditioners, including those from Indian brands.

“Alexa+ is not a chatbot or a single large language model. It has to do much more. It has to take actions in the real world. It has to be able not just to play music, but now also to book a restaurant,” Butti said.