Opening

The person who has been writing briefs for live brand activations - where the audience brings everything you cannot control, where no interface protects the brief from the room - sees something specific when the AI interface layer starts to dissolve.

What they see is not disruption. What they see is confirmation.

A live activation brief was never a navigation map. It could not be. When the user brings an unpredictable state into the space - their mood, their expectations, their prior encounter with the brand - you cannot specify what they will see at each step and trust that the interaction will hold. What you can specify is the brand's intent, the constraints that must hold regardless of what the audience brings, and what happens when the goal state cannot be reached. You brief for a system, not a sequence. You brief for outcomes and failure modes, not for screens.

This is the structural problem that the interface layer was, for most digital work, masking. The interface absorbed the unpredictability. It constrained what the user could express into a set of interactions the system could parse - the screen, the kiosk, the app, the navigation. Not a failure. A translation mechanism. And it has been true of every physical brand activation where user behaviour is mediated by a touchpoint, and of every e-commerce flow where the browse-and-checkout sequence is not the most direct path to what the user wants - it is the most direct path the system could follow.

When the interface dissolves - when an AI agent routes the user's intent directly to execution, without a predetermined navigation structure mediating what is possible - the translation mechanism changes. But the user's goal does not. And the discipline of writing briefs that hold without a fixed interface turns out to be a discipline the experiential practice has been building for decades. What changes is not the method. What changes is who needs it.

The_Interface_Was_Always_a_Compromise

This is not philosophical. It is structural.

A traditional activation brief specifies the interface layer. It says: the user sees this. Then this. Then this. The sequence of interactions is the sequence of screens. The brand signal appears at these three moments. The touchpoint does these things. The user's mental model is shaped by what the interface reveals at each step.

An intent-flow brief specifies something else entirely. It says: the user is trying to achieve this. The system must never do this, no matter how the intent is phrased. If the user's goal cannot be achieved, degrade gracefully in this way. The brand signal is held in these constraints. The experience emerges from how the system respects them, not from how the interface is laid out.

These are not the same document. The traditional brief is a navigation map. The intent-flow brief is a constraint specification and a goal-state description. You can write the first one beautifully and still produce the second one wrong - a system that navigates correctly and achieves the wrong outcome.

I have been writing both kinds of briefs in parallel. Not sequentially. Simultaneously. For different systems.

What_a_Fit-For-Purpose_Intent-Flow_Brief_Actually_Requires

When I brief my agent team - when I hand Sheena the task of orchestrating a PR review across eight repos, or task Nix with implementing a feature and shipping it without architectural debt - I am not writing a navigation map. I am writing an intent-flow specification: a system where the user's intent flows directly to execution, without a predetermined interface mediating what is possible.

I state the user's intent: what is the team actually trying to achieve, stated as an outcome rather than a task sequence?

I specify the constraint layer: what must the system never do, regardless of how the intent is expressed or how many ways the problem could be solved? When the architecture has to remain open. When shipping means zero technical debt. When a decision affects eight repositories and cannot be made by a single agent.

I define the failure mode map: what happens when intent cannot be executed? How does the system degrade gracefully? If a PR cannot be merged cleanly, do we stop and surface it, or do we escalate? If a design system change breaks backwards compatibility, do we pivot the task or do we hold the constraint? What does the user experience when the goal state is not achievable?

I specify the brand signal: what is the team expressing through this work, and how does it hold when the process is invisible? This matters more for an AI system than for a physical touchpoint, because the user is not watching the interface - they are only seeing the outcome. The brand signal is in the work itself.

The experience the user (or in this case, the team) receives is the output. Not the path to it. The interface - the conversation, the tooling, the visible scaffolding - is not the experience. It is the mechanism. The brief specifies what the experience should be. The system routes to it.

This is the same discipline. A system that executes user intent without a predetermined interface requires you to know what the brand will express regardless of the route. A live brand event where you do not control what the audience brings into the room requires you to know what the brand will express regardless of the setting. The structural problem is identical. The translation is direct.

This is also the technical dimension of the shift covered in When Agents Render the Interface - where the interface dissolves not because the brief changed but because the rendering decision itself has moved inside the agent. The governance problems converge at the same point: you cannot brief for a fixed output when the output is generated at execution time.

The teams writing intent-flow briefs now have the learning curve advantage. They are building a capability that will be non-negotiable. They are learning to hold scope fluency at a layer where the interface cannot do the work for them.

What_This_Means_for_Appointing_Brief-Writers

The traditional brief-writer is a UX designer or a producer. The skill they bring is interface literacy: they can imagine a sequence of interactions, they can see the gaps in navigation, they can translate the brand's intent into a touchpoint experience. That is a discipline. It has trained a generation of people to do it well.

The intent-flow brief-writer is a different role. They need to understand what the system is capable of, what it will infer when it does not have complete information, where it will fail, what 'the brand signal is preserved' means as a constraint specification rather than as a visual style. They need to see the architecture and the brand requirement not as separate things to be reconciled, but as the same thing. They need to understand what remains unsaid in the brief and what the system will do with it. They need to be appointed when the architecture is still open, because the brief is now the architecture.

This is scope fluency. The capacity to hold creative intent, technical requirement, production reality, and audience experience simultaneously - and see what is missing from the whole before it becomes a gap in the finished work. It has always been the capability that separates the people who change how brands actually operate from the people who execute what others have decided. It becomes non-negotiable when the interface is gone.

Three_Questions_That_Actually_Matter

If you are about to commission agentic AI work from a vendor, or bring someone in to brief it, here are the three questions that distinguish someone ready for intent-flow from someone still thinking about interfaces:

Can they specify what the system should never do, regardless of how the user's intent is expressed?

This is constraint-layer thinking. Not 'the user should see this message in this case,' but 'the system must never execute this action under these conditions, even if the user's request implies it should.' A brand-side example: the user asks the agent to book them a table at a restaurant. The agent routes to a reservation system. But the brand constraint is that the system must never book a restaurant the user has never visited - because the activation is about introducing the brand-exclusive experience, not extending it to substitutes. The agent needs that constraint encoded, not as a visual gate, but as a system boundary.

Can they describe the brand signal as a constraint, not as a visual?

The brand signal in an interface-based activation lives in the pixels - the colour, the typography, the layout, the visual hierarchy. In an intent-flow system, the signal lives in how the system respects the constraints you have set for it. The restrained choice not to guess when information is incomplete. The graceful failure that honours the brand rather than excusing itself. The speed of response that signals confidence rather than appearing effortless. These are constraint expressions, not visual expressions. The brief-writer needs to think this way.

Can they map what happens when the intent cannot be executed?

Failure mode design. What does the user experience when they ask the system to do something that violates a constraint? When they state an intent that the system cannot fulfil? When the system has to choose between the user's request and the brand constraint? A traditional brief specifies success states. An intent-flow brief specifies success states and failure states equally. It names what the system should do when it cannot do what was asked.

These questions are not new. They are the questions I ask when I am being appointed to a brief - when the architecture is still open, when the decisions are not yet locked, when what I am being asked to hold is the full shape of what is possible and what must never happen. The teams that cannot answer them today are already behind.

The_Signal_Is_In_the_Brief,_Not_the_Interface

OpenAI's Jony Ive-designed audio device - screenless, audio-first, agentic, now projected for early 2027 - is a detail. Counterpoint Research forecasts 32% agentic AI penetration in smartphones by 2027; IDC projects 60% generative AI smartphone shipment penetration by 2028. The interface dissolution these signals point to is not a future-state problem.

The interface is becoming one layer among several rather than the container that holds everything. And when the container dissolves, the container's job moves somewhere else. It moves into the brief.

The brief has always been where it starts. The interface was never the thinking - it was the output of it.