Alibaba’s AI Push: Big Potential, But Still Waiting for a Killer Use Case
This article is part of “The AI Playbook” – a series decoding how China’s tech giants are approaching AI
I’ve been watching Alibaba’s AI story unfold for a few years, and for most of that time, it’s felt oddly underwhelming.
That’s strange when you think about it. This is a company that has the most enterprise relationships in China, the largest domestic cloud infrastructure, and years of heavy R&D investment. If AI commercialization was about static advantages, Alibaba should’ve been first across the finish line.
But it wasn’t. For a long stretch, Alibaba’s AI efforts felt like fragments — one team working on a model, another launching a feature, someone else doing a PR announcement. There was no through-line, no product that stuck, no single user experience that changed how people thought about AI.
The Break Begins With Structure
What finally changed my mind about Alibaba’s AI wasn’t any technical release. It was the reorganization.
Until that point, Alibaba had been running what I’d describe as a “platform empire” — one where infrastructure, tools, and data systems were centralized, and business units drew resources from shared pools. That made sense when the goal was scale. It encouraged reuse, internal optimization, and alignment. But it also encouraged detachment from the market.
I’ve worked with enough product teams to know that if you’re not directly responsible for outcomes, you tend to optimize for internal incentives. You build systems that look great on slides, that win architecture awards, that check all the boxes in internal reviews — but don’t necessarily ship faster, convert better, or monetize smarter.
That’s what Alibaba’s AI felt like for a while. Strong architecture, little urgency.
The organizational split changed all that. Now each BU had to carry its own weight. DingTalk couldn’t rely on Alibaba Cloud to cover gaps in compute efficiency. Quark couldn’t assume brand halo would give it user loyalty. Accio, if it wanted engineering headcount, had to show impact on GMV. AI stopped being a centralized asset and started becoming a business-line responsibility.
The psychological shift that followed was more interesting than the structural one.
Teams were actually running shadow metrics on user friction. And most importantly, people stopped pitching “AI strategy decks” and started demoing “this saves five minutes per user session, here’s the usage data.”
Of course, this wasn’t perfect. Internal frictions still exist. Resource allocation is still political. But the conversation changed.
Three Products, Three Different Plays
The best way to understand Alibaba’s AI thinking right now is through three products: DingTalk, Quark, and Accio.
DingTalk is evolving from a corporate obligation into a quiet productivity engine. You don’t notice the AI — but it’s summarizing your meetings, drafting your reports, writing emails for you. More importantly, those features are tied to real pricing tiers. This isn’t free magic. It’s paid value.
Quark took a left turn I didn’t expect. Instead of trying to become a “better browser,” it went narrow and deep into education. Homework solving, essay feedback, Q&A precision. These aren’t gimmicks. They’re real, sticky use cases for students and parents. It’s not viral, but it’s trusted.
Accio might be the least visible but most commercial of the three. It’s embedded inside Alibaba’s international wholesale platform. It helps sellers write better listings, match with buyers faster, respond smarter. It’s not a product people talk about — but it’s one that helps people close deals.
Each of these is making the same bet: don’t chase hype, chase utility.
Still No Breakout
Here’s the problem: none of these products have broken through.
They’re solid. Smart. Commercially promising. But they haven’t hit that inflection point where users talk about them, share them, or demand more.
That might be okay — except competitors are starting to. ByteDance, with Doubao, is running the playbook Alibaba used to own: rapid iteration, user-first product loops, distribution at scale. Tencent is embedding AI so deeply into its products that most users don’t even realize it’s there — but they benefit from it daily.
Alibaba’s approach still feels split between ambition and execution. Like it wants to be the platform, the product, and the infrastructure — all at once. And as of now, it’s yet to fully win in any of them.
My take
If I had to summarize my view now: Alibaba doesn’t lack potential. It lacks sharpness.
It has the infrastructure. It has the clients. It has the capital and the context. What it needs now is a focused strike — one product, one moment, one story that actually lands.
DingTalk could become that story if it goes deeper into enterprise AI. Quark might win a vertical if it can own education. Accio could quietly lock in B2B sellers across China’s export machine.
But one of them needs to break out. One of them has to give the rest of the strategy a reason to exist.
Until then, Alibaba will remain the company with the most to gain in AI — and the most to prove.
-To be continued
This article is part of “The AI Playbook” – a series decoding how China’s tech giants are approaching AI.