Talking to AI is turning tools into a working team

Expanded from a LinkedIn post originally published on 2026-09-15.

I’ve realised one of the biggest upgrades to my AI workflow has nothing to do with using a smarter model.

It’s simply talking to it instead of typing.

Right now I’m building my own personal AI agent that runs on my Mac Mini and connects into things like my emails and other parts of my digital life.

At the same time, I’ll regularly be working across 3 or 4 completely different projects.

I’ll give one AI agent a task, switch projects, give another agent something to work on, then jump somewhere else while they’re both working in the background.

In one project this morning, I was effectively doing two jobs at once: debugging an issue with one agent while using another to make changes to the UI.

And voice makes this ridiculously quick.

Instead of stopping what I’m doing, carefully writing out a prompt, correcting it, and then going back to work, I can hit a key, explain what I want in normal English, and carry on.

It turns AI from something you “use” into something much closer to a team you’re constantly communicating with.

And I think that distinction is going to become increasingly important.

You don’t need to be an AI engineer to work like this either.

Developers can use it to write and debug software.

Business owners can use it for emails, research, admin and marketing.

Marketers can use it for content.

Pretty much anyone who spends a large chunk of their day on a computer can start building workflows around agents.

I’m experimenting with this stuff every day at the moment, and I’m thinking about doing some 1-to-1 sessions showing people exactly how I’ve set mine up and, more importantly, how to actually work with AI like this day-to-day.

If you want some guidance setting up your own workflow, send me a message.

Happy to talk you through it.

#AI #ArtificialIntelligence #AIAgents #Automation #Productivity #FutureOfWork

The upgrade was the interface

A smarter model is not the only way to improve a workflow. Sometimes the friction sits at the point where I have to stop working and explain what I need.

Voice lets me give the context in ordinary English, then get back to the project while the agent works. It is particularly useful when I am moving between debugging, UI changes and the other tasks that compete for attention.

Briefing still matters

Working this way does not mean every task should run unattended. Separate projects need clear briefs, and changes still need review before I trust them.

What changes is how easy it becomes to keep that communication moving. Developers can use it for code, while other computer-based work can benefit from the same pattern: explain the task, let the agent do the legwork, and check the result.