(Credit: Amazon)
Alexa+, the latest iteration of Amazon's popular AI assistant, is currently available in early access, and we expect to hear more about its new features (and maybe nail down a proper release date) at the company's Sept. 30 press event.
While the new version is replete with features, including some frequently requested options that never made it into the base release, it still lacks some core functionality that's common in other popular AI models. Here's what's currently on offer, and what I'd love to see in a future update.
What’s Available Now
In Amazon's defense, the update is already significantly more capable than the original, even in its current early access state. The biggest, overarching improvement is Alexa's conversational skills. The Plus iteration is significantly better at reading and understanding context, including remembering topics you've previously discussed and incorporating that knowledge into its responses. While the original Alexa was much more of a simple question-and-response database of information, Alexa+ feels much more on par with contemporary AI chatbots.
This includes things like voice and visual ID, which allow Alexa+ to recognize you on sight/sound and personalize its answers accordingly. You can also now continue a conversation naturally, without having to repeat the wake word every time you want to follow up or ask another question.
Alexa's memory has improved across the board. Beyond having access to previous conversations, it can also remember things like your frequent flyer number or a selection of favorite recipes you can ask for when you're struggling to figure out dinner.
(Credit: Amazon)This extends to documents, which you can provide via either direct upload or email. Alexa can scan manuals, photos, and even handwritten notes for information and provide it on the fly, or synthesize study materials and create a quiz for you.
It can also now provide more nuanced reminders. Instead of simply alerting you at a specific time/date, Alexa+ can send you a reminder when concert tickets for your favorite band go on sale, or there's a price drop on a product you've had in your cart but haven't pulled the trigger on yet. In the same vein, it can integrate your calendar and allow you to make changes to it using natural language, like "always provide an hour-long buffer between meetings." This extends to building calendar events and reminders based on email.
Speaking of email, Alexa+ can also draft and send emails and texts for you, or assist you with issues of tone or grammar when you're writing them yourself. Like a number of other LLMs, it can also assist in writing essays by providing potential topics of interest or helping you hone the flow of existing copy.
(Credit: Amazon)The list goes on. There's much greater integration with music apps for playing tracks or building playlists, the ability to search podcasts by episode title, personalized content recommendations through Fire TV, and the ability to create Routines with your voice or based on the current weather.
What I Want From Alexa+
Still, there are still a number of missing features that I hope Amazon announces next week.
1. Greater Accuracy
The biggest one is a broad improvement in accuracy. While most AI models struggle with this to some extent, the early access version of Alexa+ seems particularly bad, both in scope and frequency of inaccurate outputs.
I ran into this several times during testing. When I asked if petit fours were available to purchase online, Alexa+ first recommended rainbow cake bites, then filled puff pastries, before admitting that they weren't available to purchase and that it could only search Amazon for products.
When asked when the next Formula One Grand Prix was set to take place, Alexa repeatedly referenced previous races. It first suggested that the next race was the Australian Grand Prix, which happened in March, then went through a number of other Grand Prixs that had already taken place. This was despite listing the date the races had taken place; Alexa insisted they were "next" on the calendar each time.
(Credit: Amazon)One of the Amazon News posts about Alexa+ suggested that it could solve even vague questions about movies, with the example “What’s the song that Bradley Cooper sings at the start of that country music movie?” I tested this by asking for the theme song from the movie where Cary Elwes played the Dread Pirate Roberts, and it first believed I was talking about Casino Royale, and then Breathless. Only after repeatedly correcting and prompting did it land on the correct answer, ”Storybook Love” from The Princess Bride.
If an LLM can't provide accurate answers, it undermines a huge portion of the platform's usability. It also makes you question its accuracy in other services, like providing timely reminders or sending appropriately worded emails. Hopefully this is a priority for Amazon, because an AI assistant's credibility is essential for its efficacy (and adoption).
2. A More Platform-Agnostic Approach
Alexa is limited to strictly interacting with Amazon-owned platforms and the company's partners. This means that when shopping for most items, Alexa will only search Amazon itself and can't order items not available on the platform. There's no included capability for exploring other online storefronts or doing a more generalized web search for products, though Alexa will provide broad recommendations if it can't find an item on Amazon (like "you may want to try a speciality bakery" in the petit fours example I mentioned above).
This extends to app integration as well. Alexa+ relies on OpenTable, the online reservation system, to book restaurant reservations. This means that if a restaurant isn't on OpenTable, Alexa can't book you a reservation there. A more agentic version could directly contact a restaurant through its website, or even by phone, and make reservations regardless of whether they're associated with a specific app or service. Speaking of agentic capabilities, that's a central component of our next big feature request.
3. Broader AI Capabilities
(Credit: Amazon)Unlike a number of other current-generation AI models, Alexa+ lacks some advanced capabilities, in part due its focus on operating in an ecosystem of smart home devices. Alexa can't write code, for instance, create professional-grade images, or provide specialized expert-level knowledge in many contexts. It lacks some of the agentic properties of ChatGPT, for instance, like the ability to "see" a website, extract data, or fill out forms. This means that many complex, multi-step tasks are still beyond the latest incarnation of Alexa+.
While these may not be essential for many users, they are the kind of upgrades that would elevate Alexa+ from a useful novelty to a one-stop shop for AI services.
4. Local Processing
All interactions now require cloud processing, unlike previous iterations that allowed for some basic commands to be processed locally. This is a significant issue for users with privacy concerns.
5. Return of Voice Options and Regional Customizations
While Alexa+ can be personalized by remembering user preferences and personal details, the ability to choose an Alexa voice with a different accent has been removed, replaced with four feminine and four masculine voices. While I like the new default voice, and it definitely has more "personality" than the preceding default American voice, I do miss my posh British assistant.
A Hotbed of Potential
It may sound like I'm being quite hard on Alexa+, but to be clear, I'm fairly pleased with how far the assistant has come. Its integration with my smart home devices and the convenience of voice activation have already elevated it to one of my top AI/search go-tos. If Amazon's ambitious plans to expand and refine it proceed accordingly, I could see it fully replacing the other models I currently use in my day-to-day. We'll know more after the 30th.