Real-time collaboration: why every SaaS product is going multiplayer

Real-time collaboration used to be the thing that made Figma and Google Docs feel like magic. Steven Fabre thinks it's now table stakes. As founder of Liveblocks, he builds ready-made collaborative features, like comments, notifications, live cursors and a rich text editor, that SaaS companies drop straight into their products. And increasingly, the collaborator on the other side isn't a person. It's AI.

In this episode of Now or Never, Steven spoke with Megha Sevekari and Fra about why every SaaS product is heading towards multiplayer, what designing for human and AI collaboration looks like, and what he learned moving from designer to founder. Here's what product leaders can take from it.

Why real-time collaboration is becoming a base expectation

Steven's view is simple:

"...most SaaS software should be collaborative by default."

He compares it to what Google Docs did to Microsoft Office, and what Figma did to Sketch and InVision. Once users experience working together in the same place, a single-player tool starts to feel broken. In his words, collaboration is becoming almost a commodity, and if you don't have it, your users will drift to a competitor who does.

There's a business case too. Steven says companies using Liveblocks see users spend more time in their apps as collaborative features are added. And the growth effect is real:

"your users become your best kind of growth channel because they end up inviting new people to the product"

Someone asks a colleague what they think of a sentence, the colleague gets a notification, and suddenly there's a new user in the product.

It took InVision a year and a half. That's the problem.

Steven came to this problem as a designer of creative tools. After working on an email builder at Campaign Monitor in Sydney, he joined InVision and worked on InVision Studio, a desktop, file-based design tool. As Figma's browser-based, multiplayer approach won designers over, InVision set up a project to make Studio collaborative in the browser.

What Steven expected to take months took over a year and a half, with a full team of seven to eight people. Later, while building a presentation tool with his co-founder Guillaume, the same thing happened. Six months in, they were spending almost all their time on the real-time and multiplayer side and almost none on the tool itself.

That's when it clicked. They extracted what they'd built, turned it into APIs, and in March 2021 Liveblocks officially started. Most product teams will never have a whole team to spare for collaboration infrastructure, which is exactly the gap Steven saw.

Collaborative features are more than live cursors

Here's the nuance a lot of teams miss. When Fra asked about real time, Steven pushed back:

"It's a lot more than just real time."

Most collaboration in teams happens asynchronously. It's leaving a comment on a spreadsheet cell, or highlighting a sentence in a document and asking for it to sound more like the brand's tone of voice. It's mentioning someone and making sure they're notified at the right time without being spammed, so they come back to the product.

Being in the same file at the same moment does happen, and when it does it has to work. But it happens less often than people think. So Liveblocks grew from real-time infrastructure into the whole collaborative workflow: presence, comments, notifications and an in-app inbox.

For product teams, that's a useful prioritisation rule. If you're adding collaborative features, contextual comments and good notifications may deliver more value sooner than live cursors. Our guide to UX design for SaaS covers more on designing for real team workflows.

AI is just another collaborator

This is where Steven's thinking gets interesting. He sees Liveblocks as well positioned for AI because:

"AI is basically just another entity you need to collaborate with."

Liveblocks recently launched AI Copilots, designed to let users collaborate with AI inside a product and have it take actions there. The part Steven finds hard isn't asking the AI something. Chat, as ChatGPT showed, is a pattern people already understand, and he expects audio to matter too. The open question is what the AI does next, inside your product, and how you show it.

Borrow from how humans already work together

Steven's approach is to look at the workflows people already use with each other and ask what they look like with an AI in the room. Take a blog post. You write a draft and ask a colleague to review it. With AI, you might ask it to review the draft and have it leave comments with suggested changes, just as a human would.

The right pattern depends on the product. A text editor might suit comments and suggestions. In a canvas tool like Figma, maybe a live cursor shows what the AI is doing, though Steven notes that could get hectic when the AI works that fast. Either way, the goal is to make it obvious how to interact with the AI, without it feeling creepy.

We think this is one of the most important design problems of the next few years. When AI does more of the work, the interface has less to do and the experience of working with it carries more weight. That's the thinking behind our piece on designing AI products and why we argue the less interface you have, the more brand you need.

Sell features, not infrastructure

Liveblocks spent its first two years building a low-level real-time infrastructure layer. Talented engineers could build great things with it, but for most companies there was too much work left to get value. Steven's analogy: they'd provided the Lego blocks, but if you're not an architect, the house won't look great.

About three years in, they focused on one standard use case that suits almost any SaaS product: comments. They built the whole UI, from avatars to how a mention looks in the input, and tied it to notifications. Steven says:

"...the business really started to take off when we started shipping pre-built ready-made features"

It made positioning clearer too. Customers now arrive saying they need comments, notifications or a collaborative whiteboard, not a low-level API. And pricing improved, because:

"when you sell a feature, the perceived value is different than when you sell infrastructure."

Infrastructure gets sold on usage and needs massive scale to make meaningful money. Features can be tied to the value they create for a business, like engagement and user growth.

From builder to founder

Steven's biggest personal lesson came from realising that "a product alone won't get you a successful business." Go-to-market, distribution and positioning matter as much as the product, maybe more now.

With revenue low and investors pushing, he spent a year doing sales himself instead of hiring, on the advice of a go-to-market expert who told him founders should do each job until it works. An advisor joined his sales calls for two or three months and told him he was too nice, always giving things away but never asking for money. Steven got the business to about half a million in annual recurring revenue before hiring someone to take it over. He's now applying the same approach to marketing.

What to take from Steven's story

• Treat collaboration as a default, not a roadmap extra. Users increasingly expect it, and it can drive organic growth.
• Prioritise async collaboration. Comments and well-timed notifications often matter more than live cursors.
• Design AI as a collaborator. Start from how people already review and suggest changes, and make the AI's actions visible and easy to read.
• Package value, not plumbing. Ready-made features are easier to sell and easier to price than infrastructure.

If you're working out what collaboration or AI features belong in your product, that's the kind of problem we tackle in digital product design. And for the full story, including how Liveblocks raised its first rounds, listen to the episode with Steven Fabre on Now or Never.