On Agents

Big changes are coming for all of us. They reach the very core of how we see our work, the processes we take part in, and the professions we’ve grown used to.

Like the industrial revolution in its day, and later the arrival of the internet, I believe we’re in for a new leap in productivity — accompanied, though, by serious challenges that come with fast, large-scale change.

I’m talking about AI agents. What are they? Algorithms built on various artificial-intelligence models, wired up to other applications and databases. They talk to each other, and already today they can handle almost any task you can structure clearly.

For example, I built my own AI assistant that I talk to by voice through a Telegram bot. It recognizes my speech, manages my calendar, reads and answers my email on request, works with my contact database, sets up video calls and sends out the invitations, searches the internet for data — or just chats with me, like ChatGPT. And this is only the first step.

Right now we’re watching a real explosion of experimental uses:

— voice AI assistants with access to an entire corporate or personal database;
— travel planners that find the best hotel deals, map out routes, pick flights. With enough trust, they book it all from your card on their own and send back a detailed itinerary with a schedule and recommendations;
— agents that prepare research on a given topic, organizing what they find and attaching links to the sources;
— neural nets that analyze trends, write scripts and generate YouTube videos with viral potential;
— automated systems for finding counterparties that analyze their profiles, enter the data into a CRM and send out personalized partnership offers;
— bots that talk to your colleagues in your own voice and log the status of every task.

I’m building my own endless personal database right now — a space where I can leave audio notes about events, people, projects and experiences. At some point I’ll simply ask, “Who can help with this?” — and the system will go through my recordings, contacts and notes and surface the relevant connections. Or, say, I’ll ask, “Run an analysis on this project” — and get back a report with recommendations grounded in everything I’ve done before. There are other ideas too, which I’ll share later, once they’re built.

What’s all this leading to?

We, as “philosophers of a sort,” should already be training ourselves in a new kind of thinking — framing tasks for AI agents. The companies that learn to do this first will become the new giants — in any field.

And in the best case, we become several times more effective. And maybe we’ll finally have more time for living, for creativity, for friends and nature.

Still, I believe technology should serve people, not the other way around.

Here’s to all of us building the skill of framing our thoughts as a sequence of tasks for smart assistants! 😎