How AI agents choose brands, and what the brand team should own

Thomas Marzano in conversation with John Tiniakos-Rasmussen · Quick Takes on GenAI, Templafy · Published by Templafy on YouTube, 3 November 2026 · 45 minutes · Recorded in Copenhagen, 27 August 2026

An agent shortlists the brands it can read, then recommends the ones that fit the person: their taste, what the choice says about them, what they believe. It checks what a brand says against what others say. So the brand team's job moves upstream, to the rules the brand acts by and the knowledge it holds.

In August I sat down with John Tiniakos-Rasmussen at Templafy's office in Copenhagen for their series Quick Takes on GenAI. Below are the questions brand leaders ask me most, answered from that conversation, each with the moment in the video.

Watch the conversation on YouTube

Agentic branding is the discipline of making brands intelligible, governable, and desirable in an AI-mediated world.

How will AI agents change the way customers choose a brand?

They will find a brand on what they can read, and choose it on what they can infer about the person. An agent needs the functional facts first: what you make, what it costs, where it is available, what problem it solves. That puts you on its shortlist. A brand legible only in its specifications, though, is interchangeable the moment a competitor offers a better one. I often challenge friends and family to ask their AI for a psychographic profile of themselves, built from their conversations so far. What comes back is terrifying, and very powerful. As assistants move into the operating systems of our phones and computers, they will hold that memory and filter brands through it: your taste, what a choice says about you, what you believe. So what makes a brand lovable to people now has to be codified, so that an agent can infer it too.

“Legible to machine makes you eligible as a brand… But it's the human lovability that makes you chosen.”
Thomas Marzano, 18:40

Watch from 17:11 · Related: The Legible-Lovable Law · The Shortlist Effect

Why does AI describe our brand differently from how we describe ourselves?

Because it doesn't take the brand's word for it. An assistant reads what you publish, then looks to third parties for confirmation, because its own contract with its user is that its answers can be trusted and acted on. Where what you say and what you do part ways, it finds the difference fast. I call that the integrity gap, and the brands most exposed are the ones I call costume brands: a better face outside than what is inside. There are two honest ways to close the gap. Tone the claim down to what you deliver, or treat it as your north star and work your way up to it.

“It will go to third-party sources to see, well, is what the brand says about itself, is that actually something which the market confirms?”
Thomas Marzano, 10:20

Watch from 09:56 · Related: What happens when AI exposes the gap between what a brand says and what it does? (Legible and Lovable, Ep. 1)

How do we stay on brand when AI is producing our content?

By turning guidelines into instructions a machine can follow without having to interpret them. Guidelines and templates were built for repeatability in a chain of people. I once traced how one of our guideline PDFs travelled. An agency account manager in a region downloaded it from our portal and sent it to a freelancer they had just hired, who kept it on their laptop. Six months later the freelancer was hired again. We had updated the guidelines twenty times online; the work was still being done from that PDF. Once a brand can compose its documents and assets in close to real time, the chain disappears, and the template stops being the measure of being on brand. The guideline has to become a constitution: precise where the old document left room to interpret.

“I can be perfectly on brand and extremely off template.”
Thomas Marzano, 27:35

Watch from 24:51 · Related: Brand constitution vs brand guidelines

Who governs AI-made brand work, and what should the brand team own?

The brand team governs from upstream, which changes what control means. Its first job is the Brand Constitution: the non-negotiables and trade-offs that tell an AI agent how to act in the brand's name. Its second is what I call the Brand Brain: everything the brand knows, its standards, stories and proof points, in a governed memory the agents can query. Around the work, people govern the inputs and the outputs, curate the context the AI draws on, and architect the workflows. Review changes as well. I have built and run brand desks: an asset went to the desk, an ambiguous case went to a second line, and now and then the question reached me. That whole mechanism can now run as an agentic system, with AI monitoring how the brand is expressed and learning as it goes.

“We control by going upstream rather than doing it downstream.”
Thomas Marzano, 31:50

Watch from 30:21 · Related: The Brand Constitution · The Agentic Brand Operating System

Is AI an efficiency play, or can it grow the business?

Both, and the growth is where its potential lies. Efficiency is AI's first gift. Done well, it makes the work faster, cheaper and better, and that shows in the bottom line. Stop there and you miss the top line: the new value you can now deliver to your market, and the new ways to grow that come with it. Take the efficiencies gladly, and put your mind on the value you could not offer before.

“If you look at AI only as an efficiency, then you are missing on the true potential of AI.”
Thomas Marzano, 42:39

Watch from 42:21 · Related: How should a CMO show the value of AI when the technology isn't finished? (Legible and Lovable, Ep. 7)

How will brand management work in a few years?

As a loop that runs all the time. Today most companies work to a yearly cycle: collect the insights, make a plan, align the budgets to a list of priorities, and track them quarter by quarter. I expect brand management to become a closed agentic loop. Agents sense the market and make the signals intelligible; a moment of decision sets what to fix and what to pursue; work is orchestrated across campaigns, the website and the product; and what was learned goes back into the system, which gets more intelligent with every turn. At its heart sit the brand's constitution and what the brand knows. People step out of the execution and govern the loop. Large organisations will not get there overnight. This is the direction of travel, and the steps toward it can start now.

“Brand management will become a closed agentic loop.”
Thomas Marzano, 33:13

Watch from 33:13 · Related: The Agentic Brand Operating System

Where should we start?

Start with your market. Understand, ruthlessly, how it is reshaping what your customers will expect of your brand, and what your employees will expect too. A brand exists because of its market, and a company that looks only at its internal efficiency will miss the mark. Then separate the safe bets from the shiny objects, and take the steps that make sense in today's reality and move you toward where this is going.

“What are now the safe bets that I can make today that are not the shiny object that is going to be irrelevant tomorrow…”
Thomas Marzano, 40:43

Watch from 41:39 · Related: What should a CMO do first about AI? (Legible and Lovable, Ep. 7)

From the same day

That afternoon John and I hosted a live Templafy webinar on brand governance. Templafy's write-up, 7 ways to prepare your brand for AI agents (14 September 2026), turns the same ideas into seven steps, from replacing brand guidelines with a brand constitution to building a learning loop. The recording is on Templafy's site.

About this conversation

Quick Takes on GenAI is Templafy's interview series on generative AI at work. John Tiniakos-Rasmussen of Templafy and I recorded this episode on 27 August 2026 at Templafy's office in Copenhagen, and Templafy published it on 3 November 2026. Templafy makes software that helps organisations create business documents that stay on brand. Quotes are from the video, lightly trimmed; each timestamp opens it at that moment.

Sources

Next