AI Agents
Are consumers ready for them?
Enjoy this week’s provocation? Share it with one friend and help us grow our audience.
What is an AI agent? This is not a rhetorical question - I should be able to answer it.
I’ve been writing and talking about AI agents for over two years. I hosted agentic floor tours at CES. I cofounded an AI startup that designed and built the first collaborative AI agent for group chats and have been writing about it extensively in a separate Substack.
I am as close to the conversation about AI agents as anyone could get and even I struggle with this topic.
The problem is that I know the words that define an agent but cannot fathom their meaning. Don’t take my word for it, in 2025 McKinsey tried valiantly and ultimately produced accurate gobbledygook.
An AI agent is a software component that has the agency to act on behalf of a user or a system to perform tasks.
Helpful, right?
I kid, because the definition is not wrong as much as it is useless. You can read all you want to know about an AI agent, but like Robin Williams in Good Will Hunting, you won’t understand it until you really experience it.
And when you do - when you finally meet the unexpected outcome of an unprompted action - when you pass through the cold waft of wet stone, dry frankincense, and tourist musk, and gaze upon the Master’s fresco - you will finally understand something that no McKinsey report, marketing summit, or Substack blabbermouth can tell you about this confounding technology.
/https://tf-cmsv2-smithsonianmag-media.s3.amazonaws.com/filer/8e/53/8e53abee-28e6-42d4-8cb6-cf88d55fe10d/sistine_chapel.jpg){kind=link}
Many of you have not had that experience yet. Every time you tried, something felt wrong, and you retreated back to the conventional web. We’ve all been there, and if you want to be a part of the next internet, you should read on.
To understand what an AI agent really is, you have to start with what an AI agent really does.
PROVOCATION OF THE WEEK
Agents make mistakes.
The language of AI agents has always been superlative. As a result, we have all been primed to experience a technical innovation that boggles the mind only to be met with a confidently contrived hallucination.
And then an apology.
This is the future of technology?
Agents make mistakes because agents have been given the authority to make decisions. That is a new kind of technical paradigm that has not caught up to the hype. As more and more consumer-facing agents enter the market, they are reaching a group of viral creators that are getting a lot of traction for calling out their mistakes.
Afer all, it feels counterintuitive to herald a ground breaking technology while simultaneously correcting its spelling.
Are those creators wrong? Are they expecting something unrealistic and are we pushing an inferior product?
These questions are raging underneath the surface of the next internet. The power users I know and have met are willing to overlook the obvious mistakes as they marvel at the increased output, serendipitous epiphanies, and deeper flow states.
All this happens when you learn to let go of deterministic computation (like the calculator) and embrace probabilistic outcomes (that are occasionally, comically wrong). Doing this has the ironic effect of getting you reinvest in your own attention into your work. When you know it might be wrong, you have to double-check its output.
It’s a new paradigm, and its weird.
Do you think consumers are ready for it?
What’s Next
If you have thoughts, feedback, or a perspective worth sharing, reach out: chmiel@ondiscourse.com. You might see your reaction in next week’s edition.
ON_Member Events
ON_Discourse Group Chat
Thursday, July 16 2-3pm
Zoom
ON_Discourse Group Chats are live, high-signal discourse driven conversations with AI leaders, builders and operators.
Not your typical Zoom, these are intimate, provocative, deeply practical sessions where we sharpen each others perspectives.
ON_Discourse Group Chat: Follow My Flow
Tuesday, July 21 11am-12pm
Zoom
This is your chance to see how our members build their own AI stack and workflow in a candid, step-by-step reveal of how their stack actually works in practice.
Start a group chat with Doma
Work through your Q3 plans with your team with Doma. Admit that you have no map, no compass, and no clue, and watch you and your team chart your own path through the chaos.