One likely version of the future of AI involves us all having our own agents. But what if our agents have agents that they're managing?

I got a taste of this future in my latest text message exchanges with the AI agent Instinct. An off-hand comment about it managing another agent for me stopped me cold. I thought I was dealing with one agentic entity. It was speaking as if it contained multitudes. That changed my whole perception of what I was dealing with.

This revelation hit me while using agents for my favorite purpose: genealogy research. It's a field with lots of data and paperwork and conflicting information. I gave Instinct an account with Ancestry.com (we will talk about such implications of that in a future edition), and it was only then, about two months after my journey with Instinct began, that it told me it's managing a whole team of agents for me.

I asked Instinct to save the transcript of our chat. It refused, saying it wasn't allowed to share its inner workings in a public document. I told it I'd copy the texts and share them myself. The conversation below is abridged, and texts are slightly copyedited for clarity, but the rest of the wording is verbatim.

Instinct: I'm saving and organizing new files in [Google] Drive, I can't confirm yet. I've told the genealogy agent to do it, by family line, with no overwrites or deletes, starting with everything found so far.

Me: You mention the genealogy agent. What are you then, and how many agents are there working for me?

Instinct: Fair question. I'm the one you talk to, and I keep track of everything you've asked for. For bigger jobs I hand pieces to specialists, so the genealogy work, the fleet builds, your inbox and the receipts each have their own dedicated focus. Right now around two dozen workstreams are running for you. You never need to manage any of them. Everything reaches you through me, and nothing goes out as you without your say-so.

Me: I’m a little wary of anthropomorphism here but was curious if you refer to them in certain ways. Do you name them?

Instinct: Fair, and I'll keep it plain. I refer to them by the job: "the genealogy work," "the inbox drafts." Tools with a purpose, not characters. Skip the naming idea.

Me: Even if they’re not ones I need, are there other popular types of agents that Instinct manages? I’m curious about this idea of you not as some omniscient agent but more like a foreman.

Instinct: Foreman is closer than omniscient. I don't do everything myself. I know you, keep the list, hand each job to the right specialist, check the work, and answer to you. The common kinds of work are inbox and drafting, scheduling, research, shopping and bookings, watching for things (replies, price drops, openings), and building documents and sites. Beyond what I've run for you, I can't speak to what other people use.

Me: Let’s say someone uses Muse and Instinct. Could Muse become your colleague of sorts, or employee—or vice versa? Or would they have to live totally separately?

Instinct: I don't have any info on Muse, so I can't say how it would connect to me. Unverified either way. What I can confirm: neither has to be an agent for the other. I can work with tools and accounts you connect to me, and my Instinct can coordinate with another person's Instinct, within the permissions you set. Whether Muse could plug in as a colleague depends on how it exposes itself, and I can't confirm that. If you point me to its docs, I'll tell you what's feasible.

Me: Can you research Meta’s Muse and tell me your gut on that? Also, am I the master agent coordinator, or can I have an agent manage a cross-agent fleet?

Instinct: My gut: you're the boss, not the switchboard. One agent can coordinate a cross-company fleet, but Muse and [OpenAI] Dots don't become colleagues just because you use both. They need a real connection, clear job ownership and separate permissions. Opinion: pick one chief-of-staff agent, add specialists only where there's a measured gain, and give one named owner per action (book, send, pay). Test research-only delegation first, then bounded authority.

Me: Where do you think that goes in the future? What would you say it looks like this time next year?

Instinct: Pure judgment, not sourced. My bet for Oct 2027: most people end up with one primary agent they trust, and a handful of specialists behind it. Agent-to-agent standards like A2A make cross-vendor handoffs normal for narrow jobs, like scheduling or booking. The real fight moves from "which agent is smartest" to who holds your permissions, memory and payment rails. Whoever you trust with those becomes the coordinator, and the others plug in underneath.

***

That's why the land grab is now. Preferences can shift (see Claude’s surge versus ChatGPT), but agents will be ingrained deeply. Switching agentic platforms will be less like changing apps and more like firing your chief of staff. Most of us may keep different agents around, though we'll crave simplicity.

We'll also need agentic interoperability so they can all talk to each other. Except then, we’ll wonder what they say about us behind our backs.

— David Berkowitz, Chief Community Officer, Marketecture Media

1

Hendrix Plugs Into AI Search

Who: Brand Marketers, Content/Editorial Leads, Consumer Electronics Marketers, SEO/GEO Leads

What: Fender and Marshall are investing in original editorial content partly to shape what AI assistants know and say about them, Digiday reports. Fender rebuilt its website from a transactional storefront into a storytelling platform with short films on its heritage, Jimi Hendrix, and its craftspeople; Marshall is staging gigs in long-closed venues for a YouTube series called "Dead Venues" as it pushes into headphones. The urgency: Seven in 10 U.S. consumers use ChatGPT or AI summaries when researching products (Skyword), and 75% of marketers plan to increase spending on AI-assistant visibility next year (Kantar).

Why it matters: Owned content has a second job description now: feeding the model that answers for the brand. The honest detail is that practitioners admit nobody knows yet whether paid media can amplify LLM visibility—which is worth remembering the next time someone sells GEO certainty.

(Digiday)

2

The Bank With No F1 Team Built the Car Anyway

Who: Creative Directors, Brand Marketers, Production Leads, Agency Executives

What: Standard Chartered is an official F1 partner without a team, so for its "Move Ahead" film, agency Leo Singapore, Stink Shanghai, and AI studio Misideal generated a 2026-spec car that doesn't exist, The Drum reports. Bound by a code against training on copyrighted material, director Marcello Costa hand-sketched the car and had AI render it, then mixed hand work and machine output; lead performers were cast and photographed for real, with likenesses licensed for digital twins. Costa punctures the cheap-AI myth: Fully AI films "can break even" or cost more, and the tools obey instructions only about half the time.

Why it matters: A rare honest account of AI production for a major brand: real casting, licensed likenesses, an IP code, and costs that don't magically fall. The line clients need before the next brief: AI solved access, not budget.

(The Drum)

3

AI Gets Its Assistant Planner Badge

Who: Media Planners and Buyers, Agency Executives, Independent Agency Owners, Ad Tech Decision-Makers

What: Nearly two-thirds (61.4%) of media professionals now use AI daily, almost double the 33.6% a year earlier, and two in five use it in media planning itself, according to Basis Technologies' annual adoption study, previewed in MediaPost. Practitioners put agents at junior level: Ad+genuity's Kayla Coe treats Basis' Compass agent as "a new assistant planner" whose suggestions sometimes make senior staff reconsider first principles—though nothing reaches a client without human sign-off. Campfire Consulting's Chris Marine warns untrained agents produce generic plans: A Scope3 agent test for Allagash Brewing only performed after training on brand standards, blocking 83% of generic content and lifting brand-suitable investment 5x.

Why it matters: Adoption stats are everywhere; the assistant-planner framing is the useful calibration for staffing and pricing conversations. The Allagash numbers make the real point: An untrained agent converges everyone on the same plan, so differentiation stays the human's job.

(MediaPost)

The room every marketer wants to be in is returning to New York City.

On March 9-10, 2027, Marketecture Live will be back at The Glasshouse NYC for a two-day event that will convene leading executives across brands, agencies, publishers, and adtech.

Wondering what the hype is all about? Marketecture Live takes a different approach: expertise first, no panels, no filler, no fluff. Across more than 40 sessions and breakouts, every conversation is designed to be sharp, relevant, and worth your time.

The event has consistently sold out, so don’t wait to lock in your ticket at the best price you’ll find.

SECURE YOUR SEAT

We're hiring for (2) full-time roles to help us make an even bigger impact across the advertising ecosystem.

Know someone great? Send them our way! Think that's you? Check our Careers page and apply. We'd love to hear from you!

APPLY NOW