Skip to main content

Moving Beyond the "Assistant" Metaphor for AI

·1543 words·8 mins

Why the future of consumer AI isn’t a better “buddy,” but a better bridge.
#

2026 is shaping up to be a big year for Consumer AI

It seems to me that in 2025, enterprise AI started to develop some kind of shape. Product paradigms started to emerge for AI-as-coder, for AI-as-spreadsheet-analyst, for AI-as-audiovisual editor. We’re still in the early stages, but we’ve moved beyond the empty chat box.

Most of these enterprise paradigms conceive of AI as an assistant or copilot, which are metaphors that emphasize productivity as the goal. Today, AI is a tool to get things done.

But as we start 2026, consumer AI is still at the drawing board. The empty chat box is still there, blinking at us, waiting for what comes next… and something will certainly come next, for both software and hardware. My bet is it won’t feel like a productivity tool.

I’d like to propose this idea for 2026: that the winning teams will build consumer AI as a great social product.

Instead of focusing on productivity, perhaps instead we can conceive of AI as a social framework.

Maybe AI as a “host” is a better metaphor than an “assistant”.

This is all very abstract, so let me explain…

The social intelligence of great products
#

I’ve started to notice that when people use AI, they assume a kind of social posture, psychologically.

AI products are built to mimic human interactions, so this shouldn’t come as a surprise. When we open ChatGPT or Claude, we start a conversation. And whether we realize it or not, we bring a set of subconscious social expectations to the screen – some version of the posture that we have when talking to a stranger at a party, or asking a colleague for help.

So AI is a social product.

And from decades of experience in the tech industry, we know a bit about how to build good social products – that succeed in the market, have staying power, and generally enrich our lives.

The best social products have a kind of social intelligence. They respect the expectations we have about real-world social life, and use technology to supercharge something that wasn’t possible beforehand.

Most importantly, the best social products maximize the surface area for real human-to-human interaction. They get out of the way, they connect us to other real people. (I wrote about this in a post a few years ago, when I predicted that the Metaverse would not catch on). I think this explains the staying-power of technologies like Craigslist.

If AI were a person, what would they be like?
#

Today’s consumer AI experience is like having a (sycophantic) assistant, standing in the corner of the room, staring at you. Waiting for you to ask a question.

It’s socially awkward. And worse, it’s a failure of imagination. The “assistant” and “copilot” metaphors might work in an enterprise setting, but they’re not right for consumers.

Some companies have a vision of an alternative: AI as a companion, making AI bots that become friends and trusted advisors. But this seems like the wrong approach to me, too, because the experience will never satisfy our deep human desire for connection. AI as a companion will keep us tethered to a screen, essentially lonely, and feeling hollow at the end of the day. (I know that newspapers are constantly writing profiles of people who get married to their AI boyfriend, or spend 10 hours a day with their AI companion and think it’s great – I think these are edge cases and not the norm.)

The metaphors that inspire me more, where I think there is real material for innovation, are conceiving of consumer AI as a social facilitator, or a host.

Think about the best hosts you know—the ones who make parties feel effortless. They don’t wait for you to approach them. They meet you at the door. They introduce you to someone you’d like. They sense when you’re getting bored and redirect your attention. They bring you a drink or a bite when you need it.

An assistant is reactive, waiting for commands. A host is proactive, a master of context, especially social context.

What if AI’s job, for consumers, was not (just) to answer questions, or to edit images, but to connect people to each other and the world around them? To give them the information they need to succeed in the real world? To maximize the surface area for human-to-human interactions?

#

The talents of a host
#

I don’t have the full answer to what consumer AI should look like, in order to fully realize the technology’s potential while enriching our humanity and contributing to healthy societies.

But I do have some small ideas that I’m nurturing, many of which I’m working on at Outgoing.

One idea is about Push and something we’ve been calling Ambient Push. A great facilitator or host doesn’t wait for you to ask a question – at just the right moment, they’ll give you a suggestion, or introduce you to someone, or remind you of something you care about. This kind of push notification (regardless of whether it comes as a literal app notification, or an email, or some other kind of message) will be key to the future of Consumer AI, and getting it just right will require some nuanced design.

But there’s something in between push and pull: when you show up at the door of a party, you’re surrounded by lots of implicit context – like the weather outside, the vibe inside – so maybe before you say anything, the host offers a suggestion, or a non-verbal attitude, or they gesture in a certain direction. In a digital interface, imagine opening up an app and seeing great suggestions before you even ask a question, suggestions that respect the context you’re in. That’s the Ambient Push, and it’s key to the experience we’re developing at Outgoing.

I’m also thinking about Social Initiative. Post-COVID, our “social initiative muscle” has atrophied. Inviting people to an event, coordinating a meetup, even starting a conversation feels like a heavy lift these days. This isn’t laziness, I think, it’s more a kind of social scar tissue. We spent years learning to stay home, to be cautious, to defer connection. And those habits calcified.

What if a socially-intelligent AI could take some of the burden of making the first move, by teeing up invitations for us to send (or even inviting us to things), by putting groups together for a chat, or by proposing great connections we could make to other people?

My first startup, Aardvark, was a chatbot that let you ask any question – and instead of giving you the answer, it connected you to someone in your network who was a relevant expert. The answer was only half of the product magic, maybe the smaller half; the bigger half was actually making the connection, the social facilitation of putting two people in touch. We actually found that both people, the asker & the answerer, got real value out of the connection. I wish that kind of functionality would be built into the core of the consumer AI platforms, so that AI connected us to other real people. (Building this kind of network around an AI will also help it become sticky and have a competitive moat.)

And the thing I’m thinking about the most is how consumer AI can be aware of what’s happening in the real world. There are millions of people, every day, who are putting on performances, organizing activities, and building things for an audience IRL. If consumer AI could connect us to all of that human richness, it would start to realize its potential.

Social is the horizontal layer that AI needs
#

In the real world, a parlor trick is a great way to get attention in a crowded room. You show up with a magic trick, a funny story, or an unusual talent, and you’re remarkable – literally, worthy of a remark to a friend.

Consumer AI right now is drowning in parlor tricks, like stunning video generation, photorealistic avatars, and voice chat that feels almost-real. These are remarkable—people remark on them, and they’re spreading quickly—but remarkable isn’t the same as useful.

In a previous generation of products, Instagram built a remarkable parlor trick with photo filters, and it led to rapid early adoption. But then something surprising happened: as the platform matured (and, of course, as phone cameras got better), users stopped using the filters. The app started to have more empty space, more space for sharing real photos between people. I saw data that by 2017, the vast majority of photos on Instagram were without filters. The app became more of a social facilitator, and therefore more of a long-term utility.

A lot of people are thinking about what verticals to focus on with consumer AI – should it optimize for health, or finance, or travel? I think the most valuable innovation will be horizontal, about social facilitation. The consumer AI that wins will not replace human connection, it will create the condition for it.

At Outgoing, this is our bet. The future of AI isn’t on the screen; it’s in the real world.