Thirty-five years after Hammer warned firms not to automate the old path, most AI programs still buy another place to type.
Via Harvard Business Review: Reengineering Work: Don’t Automate, Obliterate
Digital transformation used to mean a new system of record. AI and digital transformation, as sold in 2026, usually means a chatbot bolted onto the same mess — another login, another hop, another place to re-key the date that already lived in email.
Michael Hammer wrote the diagnosis in the July–August 1990 Harvard Business Review: companies were using computers to speed up processes that did not deserve to exist. The line still lands. Most owner-led firms do not have an AI problem. They have a routing problem dressed up as a software purchase.
In a professional practice — law, accounting, advisory, the shops where a managing partner is still the index of the firm — the stream never stops. Notices, files, calls, chats, invoices. People attend to one thing at a time. The stream does not. A program that adds a dashboard in the middle of that hop does not transform anything. It taxes attention.
The firms that get a post-AI price on quality — a better file, without paying professional rates for clerical preparation — do not start with a tool. They start with how information actually moves.
What spends the day is context switching: the message, the unexpected call, then typing the same fact into a second system. The computer looks busy. The file does not get better. Herbert Simon put it in 1971: a wealth of information creates a poverty of attention. A system earns its seat only if it absorbs more than it produces.
Two kinds of attention sit on the same files. The people in the firm — partners, associates, administrative staff — and the attention a model can apply to preparation, retrieval, and follow-through. Mix those up and you get the two failures we see every week.
First: fire the person who still has to see the number. The naive cut on bookkeeping or intake is the seat. The price-efficient cut is the upload. Keep the verify. Change the job so the model prepares a feed into the tool you already run, and a person checks the posting.
Second: turn the model loose on correspondence and call it transformation. Draft mail is a high-risk output. Retrieval and preparation are not. J. C. R. Licklider’s 1960 observation still describes the work: most of the time went into getting into a position to think. Seconds to decide, once the file was in comparable form.
Deadlines often already arrive as email. The leak is the hop to a calendar someone besides the owner can see. Files already live somewhere. The leak is rebuilding the last matter from memory because the last version has no convention. Money already has a books tool. The leak is paying a professional rate to re-key what a model could have prepared.
If a key partner walked out tomorrow, how much of the operating picture walks with them? That is not a chatbot question. That is an architecture question.
Map the stream before you buy another license.
Rejected on the first pass, for a reason: AI-drafted work correspondence. A new practice platform “because we need AI.” Agents running across an unmapped drive.
They leave the existing processes intact and use computers simply to speed them up. — Michael Hammer, Harvard Business Review, July–August 1990
This is the Architecture Audit, not a software rollout.
We diagnose workflows, knowledge silos, and integration boundaries. You get a systems map you can act on with us or without us. Then, if you proceed, we engineer the knowledge layer — the firm’s institutional memory as a durable asset — and the execution work for intake, document preparation, or operational coordination. Models change. That memory stays yours.
We do not rip out the suite you already paid for. We do not invent a vendor you did not name. We do not build against an open meter.
The destination is operations: someone watching unused seats, model drift, and whether the file actually got better — Managed AI Operations, not a one-off demo. The paid entry is a Workflow ROI Audit when you want the costed roadmap first.
If your digital transformation plan is a license and a kickoff, you are automating the old path. Hammer already told you how that ends.
Ask a narrower question: where does a fact have to be typed twice? Who is the index of the firm? What would you keep if you cut the clerical preparation but not the verify?
That is AI and digital transformation that pays. Map first. Then let the model prepare. Then stamp.
This article summarizes publicly reported information and is for general informational purposes only. It does not constitute legal, tax, financial, investment, security, or compliance advice. AgentsROI.ai is not a law firm, accounting firm, or registered investment adviser. Facts, pricing, statistics, and product capabilities cited here reflect the sources listed at the time of writing and may change. Readers should verify current information independently and consult qualified professionals regarding obligations specific to their industry, jurisdiction, and circumstances—including applicable New York State and New York City requirements. AgentsROI.ai may have commercial relationships with vendors mentioned; where material, such relationships are disclosed. Nothing in this article is an endorsement of any specific AI product, model, or provider.
