Dan Bulteel

Why Agents Will Unseat More Incumbents Than Social Ever Did

by

Hey all,

15 years ago I wrote an article about the rise of a more social web for Huff Post.

At the time, Facebook was updating its News Feed and bringing Spotify, Zynga etc. into it. Google was also launching Google+. Obviously, one changed the internet. The other validated it, even if it didn't pay off.

If you think about that shift, we stopped going to the internet to browse our top 10 sites and instead Facebook became the front door to the internet as things collapsed into a feed.

It changed how we consumed news, media and entertainment, which meant it also disrupted to incumbent world order = great swell for startups.

These days I'm obsessed with this same opportunity as we move into a more agentic internet. I wrote a longer form piece on Substack I'll link at the end, but to avoid UI hopping I wanted to provide the basic summary here.

If you've been trying to build something but you're fighting against entrenched user behaviours with a competitor or feel it's just too hard to go up against a giant. Well, maybe you have a secret advantage.

Keep reading...

What's changing?

The internet was built to be:

Human-read, human-clicked, human-navigated.

Everything, like everything, has been built around this.

But now we’re introducing agents that:

  • Take our intent

  • Hold context

  • Execute tasks end-to-end

For a while, humans and agents will share the web.
But eventually, agents will do most of the surfing.

That single change breaks a lot of assumptions.

Like, what's the role of UI?

The opportunity: incumbents are optimized for old behavior

Most successful products today assume:

  • Humans initiate

  • Humans negotiate

  • Humans resolve

But in an agent-first world, the flow becomes:

Intent → agents → outcome

That breaks deeply ingrained UX patterns:

  • Links

  • Forms

  • Inboxes

  • Manual configuration

  • “Click to convert”

Which means entire categories that feel “done” suddenly aren’t.

That’s how 'immovable' incumbents get unseated - not by better features, but by behavioural mismatch.

A simple example: scheduling

Scheduling tools are incredibly efficient for humans.

But agents don’t need:

  • Booking links

  • Availability grids

  • Email threads

They just need:

Make this meeting happen, given these constraints.

So if you’re building something like @Meet-Ting, you’re no longer fighting the immense efficiency of booking links (which, to be fair, are great and undervalued tech).

You’re operating in a different world, one that’s increasingly agent-first.

This is where context matters: learning about the user, their goals, their relationships, and their ongoing preferences, not just availability slots.

With that context, agents can reason with each other far more intelligently.

What this means for builders

This is one of those shaky snow-globe moments.

Every category now has to answer:

What does this look like when agents are first-class users?

It's not theory, VISA said this is the last year we'll checkout alone, and Google, PayPal etc. all have protocols that allow agents to transact for us.

It really breaks your brain when you go down the rabbit hole.

For advertising, if there's no browsing history or clicks, how does it work? Agents with bounties? Feels like a billion dollar startup line of work alone.

Just imagine

Think about the future.

You wake up. You start your day in a LLM. You say/type what you want. Agents go do the work.

If that's the future of our online experience, so where does your product fit?

Love to hear any thoughts you have or if you're building towards this too?

Substack here: https://chiefting.substack.com/p/the-internet-was-built-for-humans

Thanks,

Dan

66 views

Add a comment

Replies

Best
Douglas Li

Agreed that UI needs to change. But I'm not bullish on going pure chat. Having to type text can be incredibly inefficient too, and LLMs are so prone to hallucinating intent incorrectly that pure chat interfaces are more likely to get things wrong instead of right. Booking a meeting at the wrong timeslot is annoying at best and unprofessional at worst.

You really want ways for the user to inspect what's going on, and confirmation controls before taking actions that can have serious consequences. Meeting scheduling is a human problem - you're dealing with calendars for people, not machines. So you still need polished, visual UI that gives people a sense of control and understanding of what's going on.

I gave a detailed talk about designing UIs in an agentic world, which goes deeper into these principles:

Dan Bulteel

@dougli Are you still based in London? Let's catch up, would love to meet properly.

You're right it doesn't work in a world where the agent can't perform the task or gets it wrong.

The way we've been approaching this: we currently have an agent connected to your calendar, operating in email + text, learning about your social ties, patterns and goals. When we get it wrong, user churns.

When we get it right, user is more than happy to delegate. In a world of delegation and trust, old UIs less efficient and voice/typing serve same purpose.

But I'd love to catch up, show you the product, and generally keep in touch for future. Loved the presentation and also dropping out of OpenAI to build is baller. Congrats!

Laurencio Trevino

This perspective is 🔥. The sift to agent first interaction changes everything it’s not about better UIs or features anymore, it’s about behavioral mismatch. Products optimized for humans will start feeling clunky, and startups that design for agents from day one could leapfrog incumbents. The scheduling example really hits home once agents can act with context and autonomy, traditional workflows feel archaic. Excited to see more builders thinking in this intent → agents → out com paradigm.

Dan Bulteel

@laurencio__trevino Loads of opportunity, breaks your brain to think about a new web, but hopefully some good assumptions and luck, and you have a new advantage!

Matt Carroll

I think there are possibly two ( or maybe even more) ways this can help create an opportunity,

  1. You can be faster to build for the agentic web you described

  2. you can completely swim the other way and just build for people who for whatever reason won’t onboard to using agents so aggressively.

    I would guess a small subset of people will probably always be in the world of “I just want a simple ___ that does ___ really well”. I think most big-cos will shove AI into anything at the cost of the user trust /lock-in they already enjoy to try and remain relevant to the main mass of users.

Good point though, I’ve never thought about what an app that isn’t designed for users at all would look like.

Dan Bulteel

@catt_marroll 100%, and I think in the opposite to social web, there's more push against AI. I'm still unsure if it's just the usual 'change is bad' or something more as it represents something much bigger that we haven't seen before (AI slop, intelligence, automation etc.).

I kinda read Google switching (or opening) to G.ai as a bit of a signal in the new home page / LLM for the web movement. And to use your analogy, I think you still get those really specific niche search engine people who don't want that and go to a privacy friendly one, a no bias engine etc. whatever they believe represents them most or just ease of use.

The future I always imagine though is Jarvis when you wake up and if you're just talking to an AI on voice, how the internet looks then too!