When Amazon debuted Alexa back in 2014, it hardly looked like a threat to Apple’s established “personal digital assistant” Siri. Four years earlier, Apple bought Siri for $200 million, debuting the iPhone 4S in 2011 as the first phone with an AI assistant. Everyone expected Siri to improve over time. Instead, Siri’s cofounders left Apple, and Siri bounced from one Apple executive to another, and then another, stumbling forward as Alexa smoothly plowed ahead.

To Apple’s partial credit, Siri hasn’t been standing still. It speaks more languages and reaches more countries than Alexa, while spanning five Apple platforms and various CarPlay vehicles. But as a user experience, Siri has a deservedly bad rep: It has continued to disappoint people and long ago became a “joke” within the AI industry. Even frequent Siri users openly admit to its many shortcomings.

Yesterday, I took a deep dive on how Amazon surpassed Siri by adding third-party Alexa skills, varied Alexa device price points, and Alexa-direct Amazon ordering abilities. If you believe that AI assistants are key to the future of computing, it’s fair to conclude that Alexa poses an existential threat to both Apple and Microsoft, which have struggled to create AI services as strong as their hardware and software. Apple could turn things around in 2018 with a three-part approach to bolstering Siri: the price of entry, quality of services, and quantity of services.

1. Lowering the price of entry with more devices

Following the debut and abrupt discontinuation of iPod Hi-Fi a decade ago, no one asked Apple to release another $350 speaker. Apple once explained that it chose that price to rest above popular Bose and Altec Lansing speakers, but below poor-selling “audiophile” speakers from companies such as Klipsch. Yet the consumer electronics industry has known for a long time that $300 is a magic number for mass adoption, and that each $50 decrease brings lots more customers to the table.

Putting aside Apple’s repeated claims about “reinventing home audio” with its speakers, the company needs products that appeal to the bulk of the market, not self-proclaimed audiophiles. And having forfeited 2016 and 2017 to Amazon, which isn’t standing still, Apple can’t wait several years to roll out additional devices. It needs to move quickly.

Apple’s decision to stop making routers suggests that the company thinks it’s above using engineering and design resources on products that won’t contribute gigantic profits immediately to its bottom line, regardless of the number of customers they attract or retain. That sort of thinking has created big opportunities for volume- rather than profit-focused companies such as Amazon.

If Apple — the world’s largest company — can’t spare the resources to make an affordable new Siri product, it could roll out the full version of Siri to its least expensive device, the $150 Apple TV. Or Apple could drop the price of an existing “full Siri” device, such as the iPod touch. But will that really work, particularly given how much Apple has downplayed these products in its lineup?

The better option would be to create a brand new “full Siri” product to hit a lower price point, and for once, let insane margins be damned. It goes without saying that Apple could sell millions of units of a $100 Siri speaker roughly akin to the popular Echo Dot, or a $150 Siri alarm clock like the Echo Spot, even while maintaining a solid premium over Amazon’s version. Just like the iPod shuffle, this sort of “inexpensive” product wouldn’t need to be Apple’s flagship — just something to get new customers in the door, and let existing customers have an affordable way to spread Apple products across their homes.

2. Improving the quality and depth of Siri services

Apple doesn’t have a choice on this one: If it doesn’t markedly improve Siri’s quality of services this year, even affordable new Siri devices won’t be able to compete with Alexa. Whether it’s due to Siri team departures, bad licensing deals, or other factors, it’s clear that bad things have been going on behind the scenes with Siri’s cloud-based “brain.” Far too often, the service is unable to respond reliably to certain requests, only in part because it’s misinterpreting words it previously understood. What’s the point of an AI assistant if you have to repeat yourself several times, then give up due to failure?

Apple did a good job in 2017 of making Siri’s voices sound smoother, and it continues to lead the AI assistant world in multilingual support. But Siri continues to have problems processing common natural language requests. You might ask it to play a song by an artist, only to hear it dial a phone contact with a similar name. Or you might ask for directions to a restaurant, then see it display map or even web results for something completely unrelated. There are so many problems with the way Siri parses requests, it’s hard to know where to begin fixing them.

The key areas that Apple needs to work on are properly directing requests for assistance to the right features, making smarter context-aware decisions about what someone likely means within an area of Siri expertise, and offering a full tree of actions a user can intuitively request at any time. When you ask Siri to read you an article from VentureBeat, it should know to go out on the web, search the site’s headlines (either through a search engine or the site itself), and offer you a chance to confirm which of several options you’d like to hear. Right now, “read me the latest article from VentureBeat” leads to the response, “Here are some tweets,” with permutations of the question leading to various comically wrong directions.

It wouldn’t hurt for Siri to have a better personality, either. Apple needs to make an executive decision about who Siri is supposed to be — friendly, somewhat sassy, or downright sarcastic — and implement it. Right now, Siri’s rare sparks of humor sound like barbs tossed by an unhappy IT guy, and no one likes an unhappy IT guy. Alexa is similarly awkward when it tosses out lame jokes, but at least it doesn’t sound bitter.

3. Increasing the quantity and diversity of Siri services

Part of Siri’s problem is correctly doing what it’s supposed to do; the other part is that it can’t handle so many things people would reasonably expect it to do. At one point, Siri seemed to be on the right track. If you asked it for historical, scientific, or math assistance, it pulled data — somewhat robotically, but generally accurately — from Wolfram Alpha. Today, Siri’s answer to the question “who’s the President of the United States” might come from a Google web search, a Wikipedia entry, or “Siri Knowledge.” On a positive note, Siri can point you to information from several sources. Unfortunately, Siri’s list of sources is short, and it frequently just points in another service’s direction, leaving you to find the information you want by yourself. This is particularly awful in a car, where you can’t read the screen.

The key to fixing this issue is forming different kinds of partnerships. For primary sources — the ones Siri taps without requiring you to install third-party apps or hooks — imagine if Apple reached out to indie services to make Siri a cultural authority. For instance, Siri might tap Genius for insights into song lyrics, Vimeo for surfacing videos, Pinterest for decorating suggestions, or Uncrate for men’s shopping trends. This would go a long way toward humanizing Siri’s output.

Second, Apple needs to get third-party developers signed up to create equivalent or superior versions of Alexa’s skills, which already boast over 25,000 options. So far, Apple’s alternative SiriKit doesn’t appear to be off to a good start on the development side, and SiriKit for HomePod is surprisingly limited — it requires sending HomePod requests to an iOS device for processing. On the user-facing side, Amazon makes adding Alexa skills easier than Apple makes downloading iMessage stickers. That should be Apple’s goal for Siri, as well.

That’s all nice, but what will Apple actually do?

Apple apologists routinely suggest that mere mortals shouldn’t dare tell Apple what to do. Despite decades of naysaying, Apple’s now the world’s largest company and has insane profit margins. From that perspective, even if Siri is deeply flawed while Alexa’s increasingly impressive, Echo speaker sales seem unlikely to hurt Apple’s sales or hugely impede its path forward.

But is that view wise for the long term? Apple once knocked industry giants IBM and Microsoft for misunderstanding the “personal” dimension in computing. They took decades to crumble in the public’s estimation, but crumble they did, making room for friendlier computers and apps created by Apple. Similarly, if Apple keeps dithering while voice-controlled, AI-assisted computing matures, while players like Amazon (and Google) keep winning over customers, the competitive landscape could soon look even worse for Siri than it does right now.

I don’t expect Apple to sit on its hands this year. History suggests that HomePod will appear in a “close but still not quite right” form by the end of March 2018 — unless Apple opts to delay it again for strategic reasons, which isn’t impossible. I would expect to hear about major improvements to both Siri and SiriKit soon before HomePod’s release, with more announcements at WWDC in June 2018.

It’s unclear whether Apple will fast-track a more affordable Siri speaker for release under its own label, or the Beats brand, by year’s end. I would love to see Apple parlay its iPod touch and Apple Watch expertise into a dedicated bedside clock like the Echo Spot, but that might be too much to expect. Apple’s development teams typically require two or more years to take a product from concept to release, so if something wasn’t already well through development last year, it’s not coming this year. An affordable iPod Touch could serve almost the same purpose without demanding design and engineering resources.

The risk to Apple is straightforward: If Siri continues to lag behind Alexa in functionality and ubiquity around the home, Apple’s competitors will continue to expand their footprints with Apple users. If there isn’t a much more affordable Siri device on the market by the end of 2018, you can expect to see Amazon and Google dominating the holiday App Store charts next year, just as they did this year. Until then, I’ll be enjoying my Echo Spot, and slowly but surely becoming more invested in Amazon’s Alexa ecosystem.