Reference
Explore
An AI Alphabet
26 letters, design history rhyming with how we use AI now.
A chair is a brief.
A factory is a prompt.
An archive is a prompt library.
The questions we are asking about AI are not novel. Our designers, factories, and archivists have already answered these, in their own medium, on their own terms. This is a working library of the rhymes, and a short list of places the rhyme breaks. MillerKnoll already knows how to do this work, in materials we have worked in for over a hundred years. Each entry names a thing we cherish and a thing we are learning to grasp.
Site content last reviewed: April 2026
The Action Office
Herman Miller, 1968
The most famous office system of the twentieth century started as a research finding, not a product. Robert Propst spent years watching knowledge workers in the field before he designed a single panel, and what he saw (people hunched over fixed desks in rooms that punished movement) made him build something that could be reconfigured on a Tuesday morning without a work order. The system had 87 components. Desks were height-adjustable. Panels came off. His phrase for the office of the future was "the vertical thinkscape," which is the kind of thing you can only say in 1968 and survive.
Every team is already assembling its own Action Office. Prompts, agents, checks, habits; a kit of parts, not a finished room. Propst would have recognized the move, because he refused to ship a room he had not watched someone try to work in.
The Bertoia Side Chair
Knoll, 1952
Harry Bertoia said his chairs were "mostly made of air, like sculpture. Space passes right through them." He welded every wire of the first prototypes himself, in a shed Hans and Florence Knoll built for him in Bally, Pennsylvania. The chair took him two years. He did it for the royalty (which Knoll paid him, and which funded the sculpture work that was his real life), and he never pretended the chair was the important thing. It was the grant.
A chair you cannot see through hides its failures. So does a model. Bertoia made the lattice visible on purpose, so you could watch the welds hold or fail in daylight. When an answer comes back from the system, the question is whether you can see the joinery or only the surface.
Cranbrook
1920s to present
Five people who changed a century walked the same small campus in the same decade. Florence Knoll, Charles Eames, Ray Kaiser, Eero Saarinen, Harry Bertoia. They all knew each other. They all borrowed from each other. Eero and Charles entered the 1940 MoMA Organic Design competition as a team and won it, because Eliel Saarinen had run the place as a studio culture, not a school. Florence called Eliel "Father Saarinen." He called her Shu. There is a photograph from 1939 of the whole group on a porch, and it looks like a band before anyone knows they are a band.
Training modules do not make designers. Rooms do. Peer learning scales in a way curriculum cannot, which is why the cohort program looks the way it does, and why the office hours are not optional decoration. The porch is the program. If you are running an AI program without one, you are running a school, and school is where the Cranbrook kids left to do the actual work.
Democratic Design
HAY, 2002
Good design should not require an architect. That was the premise Mette and Rolf Hay started with in Horsens, where Rolf had a single table and chair he had been tinkering with at home and Mette believed in them enough to build a company around them. Their first collection launched at imm Cologne in January 2003. Their first store opened on Pilestræde in central Copenhagen in 2004. They write over two hundred Christmas cards to employees each year, by hand, which is either charming or insane depending on how you feel about Christmas cards.
The flagship HAY House sits above the street with no display windows, and you find it by walking up one floor. Access is a design decision. The staircase is the thesis in brick form: good design does not come to you, you go to it, and the climb is part of the signal.
The Eames Molded Plywood Chair
Herman Miller, 1946
Charles and Ray once set fire to their own apartment trying to bend a piece of plywood. They called the rig the Kazam! machine: a bicycle pump, plaster molds, a heating element that caught on the bedroom floor. They ruined wood for three years. The chair that finally emerged, the LCW, was called "the chair of the century" by Time magazine in 1999, more than fifty years after they made it. The first ones shipped under a contract with Evans Products, not Herman Miller. George Nelson saw them and brought the Eameses to Zeeland, which is a sentence that summarizes about four different men's careers in eleven words.
First drafts are raw material, not product. A shell of plywood becomes a chair through cycles of shaping that assume the first attempt will be wrong, and the first attempt often sets the bedroom on fire. If you are treating bad outputs as failure instead of as the curriculum, you are the person who threw out the Kazam! machine after night one.
Florence Knoll's Table Settings
Knoll, 1940s to 60s
Florence sketched the table before the meal, between courses, and after. The caterers did not have to guess at her taste. The drawing carried it. She ran the Knoll Planning Unit the same way. Before selling a chair, she would draw the whole room, including where the light fell and where the conversation would happen. Her sketches were called paste-ups internally. She kept them in a flat file. Some of them are now in the MillerKnoll archive, and some of them are better briefs than most of us will write in our careers.
A prompt is not a question. It is a brief detailed enough that someone else can execute it without you in the room. Florence may have been the first person to understand what a great prompt looks like, because the caterer could not be in her head, so she drew her head onto the page.
Girard's Textile Archive
Herman Miller, 1952 to 1973
Alexander Girard did not invent each pattern from scratch. He collected. Over 100,000 folk-art objects gathered on trips to Mexico, India, Poland, and beyond; all of it mined as source material for one Herman Miller textile after another, from the day George Nelson hired him in 1952 until he stepped away in 1973. The archive was not decoration. It was his prompt library. When he needed a color, a geometry, a motif, he had already collected it. The collection is now at the Museum of International Folk Art in Santa Fe, donated by Girard himself, because he understood the whole point of a library is that somebody else uses it later.
A prompt library is an archive of things that worked. One associate figures something out. The next one does not start from a blank page. Save what worked, organize it, let the next person inherit it. The alternative is 6,000 associates each solving the same problem on a Tuesday, which is the most expensive way to run a company ever invented.
The Handoff
Every factory, every team
Every piece of MillerKnoll furniture changes hands five or six times between raw lumber and loading dock. At Portal Mill a chair crosses wood processing, upholstery, assembly, quality, logistics; each crossing is a person telling the next one what matters. A torn note on a cart. A hand-signal across a line. A scan into a system. The handoff is the spine of quality. If it breaks, nothing downstream can fix it.
A handoff is already a prompt. We have been writing instructions for each other for a century, and the only thing new is that one side of the handoff may now be a model. Which means the quiet catch that used to happen at a crossing (someone glancing at the part and saying this isn't right) has to be built back in on purpose, because the model will not glance.
Integration, not Acquisition
MillerKnoll, 2021
Coke and Pepsi finally in the same room. That is how Amy Auscherman describes taking over both archives after the 2021 Knoll acquisition, a $1.8 billion deal that Andi Owen could have branded as Herman Miller swallowing Knoll, and instead branded as MillerKnoll, with the two lineages kept distinct. Competitors for seventy years. Cranbrook classmates from the start. Now a collective.
The weak version of the AI argument is augmentation. The strong version is collective. Both voices kept, and the work is figuring out what they can only do together. Acquisition is a takeover. Integration is a partnership. One of those is a lot harder, and it is the one worth more.
Jacobsen's Series 7
Fritz Hansen, 1955
Nine million chairs sold, and the shape is only half the reason. Arne Jacobsen drew the silhouette. Søren Hansen (the founder's grandson) had spent a decade perfecting the plywood lamination technique that let a chair back curve in two planes without cracking. The 1952 Ant chair proved the method. The 1955 Series 7 made it a global standard. The shape is Jacobsen's. The chair is both of theirs. Fritz Hansen did not disappear when the architect arrived. He became the enabler. His name is still on the building.
Human intent. AI craft. Human edit. The maker does not vanish. The maker decides what the craft is for, and neither chair exists without both. Stand in the Copenhagen showroom and you will see the 1952 Ant next to current production, same gluing technique, seventy years later. The lamination is the tool. The silhouette is the point. Mixing those two up is how you end up with a chair that does not fit anybody.
Knoll's Planning Unit
Knoll, 1943 to 1971
Before Knoll sold you a chair, Florence drew you a room. The Planning Unit, which she founded in 1943 and led for two decades, was not a sales team. It was an interior architecture practice embedded inside a furniture company. Clients came in with blank floor plates. The Planning Unit gave them back a way of working. The furniture was the answer to a question the Planning Unit had asked first. Every corporate interior in post-war America was shaped, directly or indirectly, by that question.
Start with the friction, not the tool. Anyone who opens a conversation with "have you tried ChatGPT" is skipping the step Florence would have insisted on. What is the space, what do people do in it, then and only then what fills it. The floor plate comes first. If you are already buying furniture, someone else already drew your floor plate for you, and they drew it for their own room.
Lean AI
MillerKnoll, present
The Herman Miller Performance System is lean manufacturing with our accent. Genba walks. Five whys. Continuous improvement. Respect for the person doing the work. The system came from Toyota by way of the Zeeland plant floor, where associates have been refining it for decades. It is not a poster on a wall. It is a habit in the hands of a thousand people.
None of this is new. The mindset is ours already, pointed at a new tool. Remove before you add. Muscle memory over event. The teams already trained in HMPS are already trained in the discipline that makes AI work. Same habit, different line.
Manufacturing Modern
MillerKnoll Archives, 2025
Modernism is a verb. That is the argument the curators (Amy Auscherman, Alexa Hagen, Sarah Wood) made when they opened the new MillerKnoll Archives in Zeeland in June 2025 and titled the inaugural exhibition Manufacturing Modern. Modern is not a style to admire. Manufacturing is what you do when you decide the past is not enough and the future has to be made, on purpose, by someone, today.
Becoming an "AI company" is not an announcement. It is a discipline you keep practicing, the same way Herman Miller did not become modern by buying modern furniture. You manufacture it. Every day. In the work. The announcement is the easiest part, which is why so many companies stop there.
NaughtOne and British Restraint
NaughtOne, 2005
Three friends in a Yorkshire barn, trained in upholstery and product design, built a brand on understatement. A Polly chair. A Dandy table. Shapes that do not announce themselves. The joinery is visible on purpose. The materials are named on the spec sheet, not hidden. Herman Miller acquired them in 2019 and did not rename them, because renaming them would have broken the thing that made them worth acquiring.
The best prompts are NaughtOne prompts. No decoration. No performance. No "please" and "kindly." Just the brief, the audience, the constraints, the boundary of what not to include. Everything load-bearing. Nothing for show.
Open Storage
MillerKnoll Archives, 2025
Amy Auscherman's predecessor kept the collection in boxes. Amy put it on shelves. The MillerKnoll Archives are organized into three rooms: an exhibition space, a reading room, and open storage. Open storage is the one where the objects live without vitrines, where associates can walk in, where curators can pull down a prototype, where a visiting researcher can sit with a 1956 Eames Lounge the way you sit with a book.
Prompts, agents, and use cases belong on shelves, not in boxes. The prompt library is our open storage. One associate figures something out, the next does not start from scratch, and learning compounds because the shelves are accessible. The Clerkenwell showroom makes the same argument in public: three floors, Herman Miller and Knoll and Maharam sitting next to each other, touchable, with the absence of the velvet rope as the whole statement.
Propst's Research
Herman Miller, 1960s
Before Robert Propst designed a chair, he read twenty books. He was hired as director of the Herman Miller Research Corporation, not as a furniture designer, and he spent years publishing on subjects ranging from livestock identification to orthopedic fixation before he turned to the office. When he finally designed Action Office, it was backed by a book called The Office: A Facility Based on Change. Most designers shipped a chair. Propst shipped a thesis.
Research belongs before rollout. Talk to the people. Watch the friction. Twenty books, then one chair. The number of books is negotiable. The order is not.
Quiet Modernism
Eero Saarinen's Womb Chair, Knoll, 1948
Florence Knoll asked Eero Saarinen for "a basket full of pillows, something I could really curl up in." He delivered the Womb, the most unquiet-sounding name for the quietest piece of furniture in the archive. The shell is molded fiberglass. The name came from Florence. The second one ever made is in the MillerKnoll archive, because Eero gave it to his mother Loja, and it came back to Zeeland sixty years later.
An answer that shouts its intelligence is usually hiding something. An answer that does the job without announcing itself is the Womb. Florence asked for a feeling, not a spec. Saarinen delivered the feeling, and the spec followed. Most briefs skip this step, ask for the spec, and end up with a chair nobody curls up in.
Ruthless Curation
Design Within Reach, 1998
DWR grew because it said no more often than it said yes. Rob Forbes founded it in 1998 on a premise that sounded backwards at the time: you do not need every modernist piece, you need the right ones. The catalog became the discipline. What got in mattered. What stayed out mattered more. The first DWR store opened in a garage in San Francisco. Forbes left in 2004 to start Public Bikes with the same instinct: fewer choices, chosen well.
The MillerKnoll resources page is eight items for a reason. An associate who receives a list of forty AI tools receives nothing. Five tools, with a reason for each, is an AI practice. Forty is a link dump. Curation is not what you add to the catalog. It is what you refuse.
The Sketch
Every designer who ever worked for us
Every chair in the MillerKnoll archive started on a napkin. Saarinen's TWA terminal was sketched at breakfast at the Dearborn Inn. The Eameses' plywood splint was sketched on a hotel pad in the South Pacific during the war. Florence's paste-ups covered entire walls. The sketch is not the chair. It is the agreement about what the chair will be; the thing that lets the maker and the designer work apart without drifting apart.
A prompt is a sketch. The precision of the sketch determines how far the maker can run before they have to stop and check in, which means Florence's dinner drawings and Saarinen's napkin drawings and the best prompts in the library are the same practice at different scales.
The Textile as Substrate
Maharam, 1902
The thing you sit on is something the company most people cannot name. Maharam was founded in 1902 by Louis Maharam as a theatrical textile supplier on the Lower East Side, and the fabric was behind the curtain, under the set, draped on the actor. It made everything else possible and was almost never the thing you noticed. Maharam is still in the family. Michael and Stephen Maharam run it. Their Girard collection brings Alexander Girard's mid-century patterns back into production. Every MillerKnoll chair you have ever sat on likely had Maharam textile on it, and you never thought about it once, which is the entire point.
Substrate is the word we use for governance, data quality, security posture, change management. Nobody wants to talk about them. They are behind the curtain. When they are wrong, every use case on top of them fails in a way the use case cannot fix.
Unseen
D. J. De Pree, 1927
D. J. De Pree went to Herman Rummelt's house expecting to pay respects, and came home having heard poetry he did not know the man wrote. Rummelt was a millwright. He died in 1927. His wife read De Pree lines her husband had written in private, and De Pree left understanding that he had been running a company full of whole people he had barely seen. The moment became the origin story of Herman Miller's participative management ethic. Inside the company it is known as "Go to see Tracy," after the family name. It is the reason Herman Miller was one of the first American companies with an employee stock ownership plan, and the reason associates are called associates and not employees.
Adoption is a people problem, not a tool problem. The associates using AI are whole people with skills and fears and private poetry you have not read. CAMPS is a framework. "Go to see Tracy" is the instinct behind it. The VP who wants to roll out a program by memo has never been to Rummelt's house.
The Vessel
Our own frame
A sandwich in a plastic baggie, a sandwich in Tupperware, a sandwich on fine china. Three vessels. Same width. Different judgments. The right vessel is not the nicest one. It is the one that matches the task. A weekly status update on fine china does not become a better status update. A partner-ready proposal in a baggie does not survive the meeting.
Single-use prompts, reusable prompts, agents. The same judgment, applied to AI. The question is never "what is the best container" but "what does this piece of work deserve." Overbuild and the content starves. Underbuild and it leaks.
Wood Mill
Herman Miller UK, 1976
Max De Pree rated a thirty-something unknown architect 105 out of 105 and commissioned him to build the first Herman Miller UK factory. The architect was Nicholas Grimshaw. The factory was Wood Mill, on the banks of the River Avon in Bath. They called it the action factory. The modular panels came off so the building could be reconfigured as the business changed. Forty years later, Wood Mill was Grade II listed, not for what it was, but for how it was designed to keep becoming. Grimshaw went on to build the Blue Building in 1983 and Portal Mill in 2015. Three buildings. Same architect. Forty years apart.
An AI program is not a one-time deployment. The panels have to come off. Prompts, agents, training, tooling all built the way Grimshaw built Wood Mill, so next year's version does not require tearing down this year's. Watch where the L turns at Portal Mill. Wood processing flows into assembly, assembly into logistics, at the geometry of the bend. That turn is the handoff. If you have panels welded shut anywhere in your program, that is where next year will break.
Cross-brand
MillerKnoll collective
St. John's Gate was built in 1504. The Sans was built in 2019. They sit on the same block in Clerkenwell, which is where MillerKnoll London opened three floors of Herman Miller and Knoll and Maharam sharing rooms that used to be competitive walls. New ideas always lean on old bones. Knoll's planning discipline, Herman Miller's research culture, HAY's accessibility, Maharam's substrate, NaughtOne's restraint, Geiger's grain. Each contributes a thing no other brand could. The collective is the argument, and it only works if every voice keeps its accent.
Domain expertise plus AI fluency. Neither alone is enough. The best prompts come from the collision of the two, which is what the cohort program is built to stage. Five hundred years of design on one block, a priory and a showroom leaning on each other, neither pretending to be the other.
Spoken, they are identical. Written, everything changes. Y is a placeholder; twenty-fifth in a line, a variable in an equation, a shape on a map. Why is a question, maybe the oldest one. The ear cannot tell them apart. The eye cannot confuse them. Whether you land on letter or word depends entirely on where the sound sits in a sentence, who is listening, what came before.
That gap is the whole point, and it is the place the rhyme breaks. Most of this alphabet argues that our archive already answered the questions AI is asking, and most of the time that is true. Here it is not. A model is fluent in what words sound like and much weaker at what they mean in the room they were spoken in. Ask a model to transcribe "Y" and it will guess based on frequency. The word wins, because the word is more common in speech. Ask a person who was in the conversation and they will tell you instantly which one you meant, because they heard the context, not just the sound.
Florence's sketch was an act against this exact failure. The caterer could not be in her head, so she drew the context into the brief. Propst's research did the same. So did D. J. De Pree in Rummelt's kitchen, listening to poetry instead of assuming he already knew the man. The archive is full of people who refused to ship the fluent answer and kept asking until they had the right one. The model will not do that on its own. You will.
Zeeland
MillerKnoll headquarters, Michigan
Population 5,600. A. Quincy Jones designed part of the campus in the 1970s, and a covered walkway called the Spine connects manufacturing to offices, so a machinist and a VP pass each other on the way to lunch. Amy Auscherman calls the Spine her favorite piece of corporate architecture on earth. The archive is in Zeeland. The Design Yard is in Zeeland. The decision to keep the center of the company in a small Michigan town has been made, deliberately, for a hundred years.
Serious work does not require the center of the hype. The best AI work inside MillerKnoll will not look like Silicon Valley AI work. It will look like Zeeland AI work: measured, material, accountable, tied to people you see in the cafeteria. That is not a limitation. That is the whole advantage.
This is a working library. It is supposed to change. If an entry does not earn its place when you read it out loud, cut it. If a brand we love is missing, add it. Amy's archive has a million objects. This one has twenty-six. Ruthless curation, remember.
Back to the toolkit.
millerknollai.com →Review before you send, especially when the output sounds confident.
MillerKnoll AI · 2026