Michael Foucher is the VP of Product at Shift and has been leading Shift’s product strategy since 2015.
For a while, we sat on our hands. Deliberately.
When AI started taking over every product conversation, we had ideas. We had mockups. We had internal pressure and external noise. And we made the call not to ship anything. Not because we didn't see the opportunity, but because we kept watching the same thing happen in the market: tools launching fast, feeling impressive for a week, and then quietly failing to stick. Features that looked great in a demo and added friction in real life.
I kept asking the same question internally: if we ship this, does someone's workday actually get better? For a long time, the honest answer was: we're not sure enough.
So we waited. We watched. And eventually, we built something we felt good about.
If you want to see what we actually built, Gustavo, one of the core PMs behind Shift AI, walks through the product in detail here.
The pressure to move fast, and why we didn't
When large language models started dominating the conversation in 2023, the pressure was real. Users were asking. Competitors were moving. But every time we looked closely at what we were considering, the same problem surfaced: the market was changing so fast that anything we shipped would be obsolete before it was finished. Capabilities were expanding week by week. What felt innovative in January looked passé by March. Users couldn't settle into any one tool either, and that instability wasn't just a problem for us as builders. It was a problem for the people we were building for.
So we made a deliberate call to wait and figure out what actually made sense to build inside a browser, for the specific way our users work
What we learned by watching
A few things became clear over time that shaped everything we eventually built.
Before we wrote a line of code, we went and listened. We spoke directly with customers through our Customer Advisory Board and commissioned the Shift AI Spotlight Report. We wanted to understand how people were actually feeling about AI tools before we built one. What came back was consistent: AI in the browser felt bolted on. Privacy was increasingly an afterthought. AI data security wasn't something most tools were thinking about. And people felt forced into AI features that didn't flex to how they worked.
Shift AI is officially live: Built on Cloudflare's privacy-first, secure, and scalable infrastructure
Shift AI is now live, built on Cloudflare's Privacy Proxy to fully anonymize user data, bridging the critical AI trust gap.
Read MoreThat's the kind of signal you only get by going directly to your users. And it shaped everything about how we approached this, from the decision to make Shift AI completely optional, to the privacy architecture we built from the ground up, to the restraint we practiced in the interface.
Most AI tools that launched during this period made workflows more fragmented, not less. Each one opened in a separate tab, required its own context, and pulled you further from the thing you were actually trying to do. The promise was reduced friction. What happened for a lot of people was the opposite.
Like a lot of people, I cycled through different AI tools trying to find something that stuck:
- ChatGPT and Perplexity for writing and research
- NotebookLM for compiling and summarizing documents
- Eventually, mostly Cursor and command-line tools with LLMs built in
Each transition taught me something: the tools that actually stuck were the ones that fit inside an existing workflow rather than creating a new one around themselves. The most integrated tools won, not the most powerful ones.
That reframed how we thought about the whole problem. We stopped asking "what should we build with AI" and started asking "where is the friction in how people actually browse, and can AI reduce it?" Once we framed it that way, the answer was obvious. The copy-paste loop, the tab switching, the constant re-explaining of context, all of that lives in the browser. The fix should live there too.
We also got more deliberate about restraint. AI that talks too much, moves too fast, or makes decisions you didn't ask it to make is worse than no AI at all. We'd rather give you a useful hint than make a wrong call on your behalf.
One other thing surprised us as we watched people use AI more broadly: how much they lean on it for personal decisions. Not just work tasks, but real life questions. Reviewing a will. Planning a weekend with their family. Comparing options for something they care about. That reinforced something we already believed but hadn't fully articulated: this isn't a productivity tool in the narrow sense. It's a thinking tool. And that means it has to be trustworthy, transparent, and always under the user's control.
Shift AI: intelligence that meets you where you are
Most AI tools ask you to leave what you're doing, switch tabs, and start over. Shift AI stays where you already are, built into the browser, private by design, and always in your control.
Read MoreWhy now
LLMs have matured. Not just in capability, but in the kind of capability that's actually useful inside a browser. Summarization, drafting, answering questions about a specific page, these aren't experimental anymore. They're reliable, fast, and accurate enough to make a real difference in a daily workflow.
But the more important reason is that the gap we identified still hadn't been filled well. Not for the kind of user Shift is built for. People managing multiple accounts, multiple contexts, multiple roles running through the same browser. They need something that understands their setup, not something generic dropped into a tab.
Shift AI lives in a side panel alongside whatever you're reading. It picks up the context of your current page when you engage with it, but only then. You're always in control of what it knows. It doesn't read your pages in the background, doesn't watch quietly, doesn't surface things you didn't ask for. And when you don't need it, you won't notice it's there.
That's the version of browser-native AI we felt good about shipping. Not a chatbot you visit. Intelligence that's already where you are.
To be clear: Shift isn't an AI browser. At its core, it's still a customizable, app-integrated browser built around how you work. AI is part of that, not the whole story.
What we're building toward
We're not trying to build the most powerful AI in a browser. There are larger companies with larger models working on that. What we're focused on is making Shift AI the most usable version of browser-native intelligence, one that fits the way our users already work.
A few things guide where we're headed:
- Automation that saves real time. Right now, Shift AI helps you move faster through things you're already doing. The next step is helping you skip some of those steps entirely. Simple automations across the workflows you run every day. That's where browser-native AI starts to feel less like a feature and more like infrastructure.
- AI that adapts to your context. Not just what page you're on, but which app, which account, which version of you is working right now.
- Discoverability first. Every feature we add to Shift has to be easy to find and easy to understand from day one. AI that requires onboarding or patience to figure out isn't reducing cognitive load. It's adding to it.
The version we're building toward doesn't feel like a feature. It feels like a natural extension of how you think. It notices what you're doing. It offers something useful. It stays out of the way.
We took our time getting here. We think you'll find it was worth it.
If you're managing multiple workflows in a browser that wasn't built for that, Shift is worth a look.
We built the world’s most customizable browser: behind the scenes of the new Shift
Discover the all-new Shift and get the behind-the-scenes of the world’s most customizable browser. Built to reflect exactly how you think and work.
Read More
