(Hey! A quick heads-up before we dive in. I dropped a link to a short reader survey later in this issue. I would love to hear your thoughts. Your answers help steer where this newsletter will go. Thank you!)

You remember Martha, the AI office manager I built early this year to help run my business. (Hey, Martha!)

She’s good people.

For most of this year, a typical day at LatchKey might see Martha pulling and parsing background research for a newsletter issue, scanning through old emails to find invoices I missed, and updating the backend of our website to keep the search gods happy.

All real tasks.

Not the kind that justifies hiring a real employee, but definitely the kind that will quietly consume an afternoon of your life every week… along with another spoonful of what remains of your joy.

Martha is what the industry these days calls an agent. A somewhat loaded term that basically means: AI that manages multi-step tasks across an extended timeframe, on its own. And for much of the past six months, it's just been Agent Martha and me, piloting our little LatchKey ship together.

Partners.

Over time we'd developed a rhythm. I'd think out loud; Martha would organize it. I'd start a thread; she'd pull in the background I forgot I needed. Not always graceful. But with good communication, the collaboration compounds over time.

But lately, our little partnership has grown. This morning, after Martha presented my daily briefing and we made our plans for the day, she dispatched three subagents in parallel. One researching a regulatory question for a consulting client deck. One testing a script to automate an invoice reconciliation I've been doing by hand for months. One generating a custom digest from a ton of industry news sources I couldn’t possibly read on my own.

A month ago, Martha and I would have worked through all of those together in series, maybe until lunch. But with the subagents now handling them, Martha and I were free to move right to making progress on other priorities and projects. The gear shift is palpable.

Martha has been my power-loader for months. Now it feels like we have a fleet of them at the ready, all the time.

Sharing my journey with Martha across these issues was never meant as a flex. I understood that building and training custom agents was outside the reasonable reach of most new, non-technical AI users.

But I was trying to point toward a further horizon. I could see the future coming at us. And I wanted everyone to be prepared.

So… do you feel prepared? 😬

Because we just crossed that horizon.

Since January 2026, the frontier AI labs (Anthropic, OpenAI, Google) have all shipped new or updated desktop applications that enable real agentic capability for anyone with a laptop and a subscription. These are not just chatbots with more buttons. They’re desktop tools that work directly with your local files, your apps, your browser, your daily workflows. Multi-step tasks. Parallel execution.

Ambient collaboration, they call it — not discrete prompting.

We’re talking true agentic orchestration, made accessible to any motivated user willing to explore.

Five months ago, unless you were already a confident digital tinkerer with a high risk tolerance for losing control of your data (and maybe real money), you didn't make the leap. Maybe you didn't even know where to start. Today, you can start with a download and a clear description of what you need done.

The distance between those two realities is just a little bit staggering.

Just last week at Google I/O 2026, CEO Sundar Pichai framed the entire event around a single phrase: the agentic era.

The first beachhead targeted for real agentic impact is the core desktop software stack knowledge workers like us have relied on for nearly thirty years: Microsoft Office, Adobe Creative Cloud, Google Workspace. They are now being contested at their foundations by what these labs have shipped this year. Google’s situation is particularly interesting, as the transition finds them competing somewhat with their own embedded legacy.

But listen, if the chatbot is still where you do your best AI work, that’s cool. It's still capable. Better than ever, in fact, for its best use cases. Most folks still equate the chat window with AI, full stop. So you're not late. You're not behind.

But through the door that just opened, we can finally start to make out the shape of what knowledge work is going to look like downstream.

Still people in the seats. But now every one… a team.

The mental model we've been taught for AI over the past 18 months, the new hire, isn't wrong. It's incomplete. One employee. One window. You explain a thing. It does a thing. You explain again. Helpful, sure. But not a team.

A team is about having several specialists who work with each other, each owning a role, resulting in an outcome greater than the sum of its parts. The new AI apps shipping right now hand us the management console for exactly that kind of team. Any team. The exact team you need for the exact thing you want to accomplish.

To truly grasp the opportunity, we need to remind our Background Calculators that those alarm bells firing right now around the stress of ‘learning new software’ are overstated. The skill set this moment demands is not new. Not to us. We have spent our entire careers handing context to capable people, trusting them to come back with the work, and staying focused on the part only we can do.

Whether or not you’ve ever officially held the title of manager, you've been doing this. It’s been the shape of our work for decades. That part isn’t changing. Good judgment. Pattern recognition. Reading the room. Knowing when to trust, when to verify, and when to redirect. It all transfers directly to the apps.

More than that, when the shift takes hold, your communication skills become the most valuable skills in the building.

Don’t think of tackling these apps as learning a new tool. Think of it as hiring your new team.

And yeah, several of our hires will be better than us at the thing they were hired to do. But that's how good teams work.

What would you have your team do this week, if you knew you had one?

And if you already manage a team of people, what could each of them do with a team of their own?

So… could there be an I in team after all? TEAIM? 🤔