The Design
Intelligence Review
Sign in Join free
The Agentic Design Stack
UX is the future of AI
1.2M+
People in the UXM Community
112+
Invisible Machines episodes
20+
Years publishing UX thinking
Season 7—Spring 2026  ·  17 episodes  ·  Through-line: the human architecture of institutional change
Boring AI
is Good AI
Gordon
Why AI Scaffolding
Matters More than Use Cases
Flowers · S6
Decentralized
AI is the Future
Goertzel
Canonical
Knowledge is Key
DosSantos
What AI as
Cheap Prediction Means
Gans
Cheap Prediction,
Expensive Change
Goldfarb
Folk Law &
Fax Machines
Pahlka
Inside The
Infinity Machine
Mallaby
Crisis Is
Your Opening
Nitze
The Confabulation
Machine
Ratliff
No Strategy
Without Vision
Evergreen
When Agents Have
Wallets, Trust Is Currency
Cohen Freue
Nuclear Fusion,
No Power Lines
Frankle
The Checklist Your
Deck Is Missing
McMillan
Knowing Before
Doing
Hasbe
The Pyramid Was
Always Wrong
Barzun
Structure Before
Features
Schwartz
Selling to
Machines
Scheibenreif
What Agentic
Orchestration Actually Means
Hosts ← Latest
New Ideation
New this week  ·  S7E17 · Hosts
Ideation

What Agentic Orchestration Actually Means

The demo works. People start depending on it. Then one morning you have to explain yesterday’s decision to someone who wasn’t in the room. That is the moment most “agentic” projects discover they were never orchestrated. They were demos. Josh Tyson and Robb Wilson take the terms the market glued together—agentic orchestration—and decompose the contradiction: vendors map how calls flow; the missing axis is who is allowed to decide.

Originating episode · What Agentic Orchestration Actually Means · Invisible Machines  ·  Josh Tyson & Robb Wilson  ·  S7E17  ·  September 2026
Read the full Ideation →

You cannot trust the orchestratee to be the orchestrator.

—Robb Wilson
Related ideas: Agentic orchestration Execution envelope Smoothie analogy Accountability Systems design
In this ideation
Episode
Free
What Agentic Orchestration Actually Means—Josh Tyson & Robb Wilson · Invisible Machines
Josh Tyson & Robb Wilson  ·  Hosts, Invisible Machines
What Agentic Orchestration Actually Means · Invisible Machines
S7E17  ·  September 2026  ·  Video + transcript
UXM Article  ·  From the Practice
Free  ·  5 min
Robb Wilson
What an Agentic Runtime Environment Is—and Isn’t
The execution layer that runs agent loops in production—orchestrated, governed, and bounded.
Read article →
UXM Article  ·  From the Practice
Free  ·  8 min
Robb Wilson
A Brief History of Agent Runtimes
From developer runtimes to the Dreamliner’s critical/non-critical split, and why an execution envelope beats guardrails.
Read article →
UXM Article  ·  From the Archive
Free  ·  8 min
Robb Wilson
You Can Automate a 787 — You Can Automate a Company
The Dreamliner cockpit, conversational AI, and why automating a company is the same design problem at enterprise scale.
Read article →

New Articles

All articles →
UX × AI Series
Part 10 · Organizational design

Who Owns the Design When AI Made Half of It?

Tushar Deshmukh closes the UX × AI series on authorship, copyright, and what ownership means when a design is half human judgment and half machine generation.

Tushar Deshmukh  ·  17 min
Service design

You Don’t Have to Ship It to Everyone

The quiet freedom of designing for some, not all. Agile doesn’t mean releasing everything at once—good teams roll out in layers.

Paivi Salminen  ·  3 min
Interaction design

UX is Design: Welcome Ally

Accessibility isn’t a feature for a niche. Ally is the default persona of human reality—and the foundation of AI-adaptive interfaces.

Krystian M. Frahn  ·  10 min
Conversation design

Your Users Know It’s a Machine. They’re Bonding With It Anyway.

A Communications Psychology study finds people bond with AI that mirrors them—even when told nobody is home. Disclosure is not the guardrail.

UX Magazine Staff  ·  7 min
Organizational design

Token Maxing Is a Lazy Metric

Don Scheibenreif on token maxing as a lazy metric—bottom-up adoption, top-down denial, and the difference between burning tokens and reinventing work.

UX Magazine Staff  ·  5 min
 UX × AI Series
Part 9 · Interaction design

Agentic UX: When the Interface Stops Waiting for You

Tushar Deshmukh on the design challenge of agentic systems—planning visibility, the autonomy slider, and why undo is the most important feature when the interface stops waiting for you.

Tushar Deshmukh  ·  17 min
 Members Essay
Conversation design · 🔒 Members

Usurper, My Collaborator

Josh Tyson on feeding a finished novel to Claude, and why collaborating with the usurper is easier than standing watch.

Josh Tyson  ·  10 min
 UX × AI Series
Part 8 · Interaction design

The UX of AI Itself Is Broken

Tushar Deshmukh on why AI products keep failing usability heuristics the profession settled three decades ago—and what one visible error does to user trust.

Tushar Deshmukh  ·  19 min
Ideations—Season 7 All ideations →
Future-Solving
Naming the human-feeling outcome before the tool parade begins: map what must become true, let exploration stay fuzzy while the vision stays sharp, and refuse to substitute vendor matrices for strategy.
↳ Evergreen, S7E9
Also in: vendor bake-off essays, members kickflip arc
◉ Episode ◉ UXM article ◉ Members essay ◉ Transcript
Canonical Knowledge
One governed layer of meaning everyone agrees to treat as source-of-truth—so agents and humans aren’t improvising incompatible copies of the same policy, pricing rule, or clinical pathway.
↳ DosSantos, S7E2
Also in: scaffolding episodes on operational truth
◉ Episode ◉ Guest article ◉ Transcript ○ Members essay
Cheap Prediction, Expensive Judgment
When prediction drops toward zero marginal cost, the binding constraint stops being “can the model guess right?” and becomes “who is accountable for acting on it inside messy institutions?”
↳ Goldfarb & Gans, S7E3–4
Also in: complements to prediction economics
◉ Episode ◉ UXM article ◉ Transcript ○ Members essay
The Confabulation Machine
LLMs as conviction engines: fluent, plausible, and prone to completing narratives that never happened—a design problem for verification layers, not just model knobs.
↳ Ratliff, S7
Also in: Shell Game cluster essays
◉ Episode ◉ UXM essays ◉ Members essay ◉ Transcript
Latest Episode Episode page →
Read now  ·  Invisible Machines S7E17
What Agentic Orchestration Actually Means—Josh Tyson & Robb Wilson · Invisible Machines S7E17
Josh Tyson & Robb Wilson on agentic orchestration
Josh Tyson & Robb Wilson  ·  Hosts, Invisible Machines
The demo works until you have to explain yesterday’s decision. Why mission-critical agentic orchestration only sounds like an oxymoron—and why you cannot trust the orchestratee to be the orchestrator.
Recent Episodes
S7 · Don Scheibenreif
Selling to Machines
Adoption is bottom-up while denial is top-down.
S7 · Evan J. Schwartz
Structure Before Features
You own the outcome, AI owns the output—structure before the feature.
S7 · Matthew Barzun
The Pyramid Was Always Wrong
Bottom-up is still a pyramid—constellation thinking for agentic orgs.
S7 · Sudhir Hasbe
Knowing Before Doing
Graph databases as the agent knowledge layer—knowing before doing.
S7 · Jeff McMillan
The Checklist Your Deck Is Missing
Knowledge before agents at enterprise scale.
S7 · Jonathan Frankle
Nuclear Fusion, No Power Lines
Fusion-grade models without the grid—and specification beyond benchmarks.
S7 · Brian Evergreen
No Strategy Without Vision
Vision-first strategy and procurement theater.
S7 · Federico Cohen Freue
When Agents Have Wallets, Trust Is Currency
Knowledge before action at enterprise scale.
S7 · Evan Ratliff
The Confabulation Machine
Truth, narrative completion, and verification design.
S7 · Marina Nitze
Crisis Is Your Opening
Sensemaking failure and crisis engineering practice.
S7 · Sebastian Mallaby
Inside The Infinity Machine
Capital, concentration, and frontier models.
S7 · Jennifer Pahlka
Friction Is the Feature
Folk law, implementation gaps, and civic systems.
S7 · Avi Goldfarb
Cheap Prediction, Expensive Change
Complements, judgment, and organizational cost stacks.
Members Essays Member essays →
Members only  ·  Josh Tyson
Usurper, My Collaborator

On feeding a finished novel to Claude, and why collaborating with the usurper is easier than standing watch.

Original analysis  ·  Tied to Selling to Machines
Selling to Machines Collaboration Writing Claude
Josh Tyson  ·  10 min read  ·  August 20, 2026
Read members essay →
Also in members
Knowing Before Doing Canonical knowledge
How Do You Like Them Apples?
🔒 Members essay · Josh Tyson · July 14, 2026 · 8 min
Invisible Machines Social contract
Did We Agree to This?
🔒 Members essay · Josh Tyson · May 6, 2026 · 9 min