Miyake 1819 All articles
Strategy & Leadership

Let the Software Count the Nails: A Two-Century Workshop's Guide to AI

Miyake 1819
Let the Software Count the Nails: A Two-Century Workshop's Guide to AI

Photo: Slyronit, CC BY-SA 4.0, via Wikimedia Commons

In the workshop's oldest surviving ledger, a bound volume begun in 1819 and still consulted when a question about an old commission arises, the entries are almost entirely unremarkable. Charcoal. A hinge repaired. The name of a customer's second son. Rice.

What the ledger does not contain is anything that made the work worth commissioning: which grain the master set aside, why a particular customer was greeted differently on a cold morning, or the two hours spent not talking before a price was named.

That absence is the most useful thing a two-century-old workshop can tell an American owner in 2026. The ledger was always a candidate for automation. The hand was not.

The False Choice in Every AI Conversation

Most of the advice now reaching Main Street arrives in one of two shapes. The first says that a small business which does not adopt AI this year will be outpriced by one that does. The second says that craft, judgment, and personal relationships are precisely what software cannot replace, so an owner should proceed carefully and change very little.

Both are half true, and the half that is wrong is expensive. The first ignores what customers actually pay for. The second ignores how much of an owner's week is spent on work no customer ever sees, values, or remembers.

The ledger offers a cleaner distinction. Ask of any task: is this a record of what happened, or is it the thing that happened? Records can be kept by anything. Judgment has to be kept by someone.

What a Workshop Would Automate First

Set aside, for a moment, everything ambitious. The lowest-risk, highest-return uses of AI in an American small business are almost embarrassingly administrative.

None of these functions requires the software to have taste. All of them return hours to the person who does.

Three Red Lines

Before adopting any tool, an owner should write down what it will never be permitted to do. Three lines hold up well under pressure.

The first: no machine speaks the first sentence of a relationship. The greeting, the introduction, the acknowledgment of a referral. These are short, they do not scale, and they are the entire reason a client tells someone else about you. Delegating them is not efficiency. It is a decision to be forgettable.

The second: no machine delivers bad news. A delay, a price increase, a mistake, a refusal. Clients forgive hard news delivered by a person who owns it. They do not forgive hard news that arrives in the voice of a template.

The third: no machine is allowed to explain why. Why the price is what it is. Why the timeline runs longer than a competitor's. Why this recommendation rather than another. Explanation is where expertise becomes visible, and expertise that cannot be explained reads as guesswork.

The Ninety-Day Sequence

Owners who want a practical order of operations can borrow the workshop's apprenticeship logic: learn the small motions before the large ones.

Days one through thirty. Choose one administrative function and automate it completely. Follow-up and scheduling are the usual candidates. Measure hours returned, not software installed.

Days thirty-one through sixty. Put the returned hours somewhere visible. A workshop does not bank saved time. It spends it at the bench. If the hours disappear into email, the exercise has failed regardless of the tool.

Days sixty-one through ninety. Teach one other person what you learned, in writing. The workshop's continuity came from records that outlived the people who made them. An AI policy that lives only in the owner's head will not outlive the owner's attention span.

What the Hand Still Has to Do

There is a temptation, once the administrative layer is automated, to keep going: to let the software draft the client note, propose the recommendation, summarize the conversation you did not attend. This is where the returns turn negative, quietly and without a warning light.

The ledger survived two centuries because it recorded only what it could record accurately and left the rest to be carried by people trained to carry it. That division of labor is not nostalgia. It is the reason a customer in 1819 and a customer today were served by the same institution rather than by two unrelated ones.

Automate the record. Protect the hand. An owner who can say plainly which is which has already made the most consequential AI decision of the year.

All Articles

Related Articles

One Dish, Two Centuries: The Case for a Narrower Strategy

One Dish, Two Centuries: The Case for a Narrower Strategy

What Happens After the Handshake: Building the Second Meeting American Business Keeps Skipping

What Happens After the Handshake: Building the Second Meeting American Business Keeps Skipping

Three Names on One Account: The Japanese Banking Tradition American Community Lenders Are Quietly Reviving

Three Names on One Account: The Japanese Banking Tradition American Community Lenders Are Quietly Reviving