The Return of the Assistant

The DIY knowledge worker made sense for forty years. AI assistants offer something better: the freedom of DIY with the leverage of delegation.

In the 1950s and 1960s, every executive had a secretary. Typing pools handled document production. Stenographers transcribed dictation. The professional focused on content; the secretary handled execution.

This wasn’t inefficiency. It was division of labor. The executive’s time was too valuable to spend on formatting, typing, and filing. The secretary’s specialized skills—shorthand, typing speed, knowledge of office systems—meant execution was faster and more reliable than if the executive did it themselves.

Then technology arrived, and we made a rational choice.

The Liberation of DIY

The personal computer was sold as liberation—and it delivered.

Word processing eliminated the typing pool. Spreadsheets replaced the accounting department’s adding machines. Email displaced the mail room. The knowledge worker could now do everything themselves.

The cultural resistance was real. Male executives were wary of keyboards, which they regarded as feminine instruments. Computer manufacturers had to reconceptualize the keyboard—which had been a feminine device for most of its history—into part of the new masculine computer. Typing Tutor became a bestseller because executives literally couldn’t type.

But the benefits were genuine. A skilled secretary in 1960 earned roughly $200 per week—about $108,000 annually in today’s dollars. A PC cost a fraction of that. And beyond the economics, the PC removed real coordination overhead.

Two people working in the same office on the same nominal schedule—9 to 5—are only actually available to each other about 80% of the time, once you account for sick days, vacation, lunch, meetings, and other interruptions. The secretary had their own priorities, their own workflow, their own queue. Your urgent task waited behind someone else’s urgent task—and behind your own other urgent tasks competing for the same limited resource. And if you were in a meeting or getting coffee, you couldn’t even hand off the work. The coordination overhead was multiplicative.

With DIY productivity tools, the knowledge worker could work without an intermediary. Any time of day. Any day of the week. On their own schedule, with their own priorities. No waiting. No coordination overhead. No dependency on another person’s availability.

But the freedom came with a cost we absorbed and normalized: the execution burden shifted to the professional. The secretary’s job didn’t disappear. It was absorbed. Every knowledge worker became their own secretary—and most of us are bad at it.

The Hidden Tax

The productivity loss adds up.

APQC research finds that knowledge workers spend only 30 productive hours in a 40-hour week. The rest disappears into overhead: 3.6 hours weekly managing internal communication, 2.8 hours searching for information, 2.2 hours in unnecessary meetings.

Workers spend 25-35% of their time just looking for information they need to do their jobs.

One estimate suggests $250 million is wasted daily on poor PowerPoint presentations—time spent wrestling with animations, formatting, and layouts rather than thinking about content and impact.

This is the hidden tax of the DIY knowledge worker. We format instead of think. We wrestle with tools instead of focusing on expertise. We’ve become amateur secretaries to ourselves, and amateurs are slow.

The insight from The Real Job of Support Teams applies directly: “Wasted overhead time is support’s finger on the scale of the work/life balance.” The secretary kept that finger off the scale. The DIY era put it back on.

The Intermediary’s Defense

The DIY era didn’t just shift burden—it created an industry. Microsoft, Adobe, MindManager, and the entire “productivity software” ecosystem exist because professionals needed tools to do their own secretarial work.

This industry has a vested interest in the DIY model’s survival.

Consider Microsoft’s response to AI: Copilot is embedded inside Word, Excel, and PowerPoint. Not Copilot instead of Word. The AI helps you use their tools more effectively. It doesn’t eliminate the need for their tools.

This is deliberate. If you could describe a document in natural language and have it generated directly, you wouldn’t need Word. If you could specify an analysis and have code produce it, you wouldn’t need Excel. If you could articulate a presentation’s structure and have it materialize, you wouldn’t need PowerPoint.

The productivity software industry’s response to AI is to embed it within their intermediary layer, not to let it bypass that layer entirely. Copilot in Word preserves the subscription. Copilot instead of Word threatens it.

Automating Expertise Gets Easier identifies this dynamic: “Software as product faces pressure… If the interface is generated on demand and business logic is expressed in plain language specifications, what exactly is the product?” The productivity software industry knows the answer. They’re protecting the intermediary layer.

Document formats remain deliberately complex. Programmatic access to Office documents requires specialized libraries and arcane knowledge. The complexity that once justified secretarial expertise now justifies software subscriptions. The intermediary has changed form, not function.

The Return of the Assistant

AI assistants return us to the division of labor that worked.

The professional describes what they want. The AI handles execution. The professional reviews the output before it goes anywhere. This is exactly how the secretary model operated—dictate, transcribe, review, send—but with a more capable executor.

Agentic AI as Universal Interface describes why this works now: “Two barriers have prevented most expertise from becoming software: coding skill and system access knowledge. Agentic AI removes both.”

The secretary had specialized skills the executive lacked—shorthand, typing, knowledge of office systems. Those skills were barriers to direct execution. The DIY era forced professionals to acquire those skills themselves (or suffer the productivity tax of not having them). AI removes the barriers entirely.

The AI assistant is:

• More capable than any secretary. It can write code, analyze data, draft in any style, translate between formats. A 1960s secretary couldn’t build you a financial model. An AI assistant can.

• More capable than the professional at execution tasks. Most of us are mediocre at formatting, formula construction, and slide design. We learned these skills reluctantly and practice them occasionally. The AI executes them reliably.

• Faster. No scheduling, no back-and-forth, no waiting for the typing pool.

The economics that killed the secretary now favor the AI assistant. The execution burden can shift back off the professional’s shoulders.

Reviewability: Why This Works

The secretary model worked because of a built-in verification step. You dictated a letter; the secretary typed it; you reviewed it before signing. The professional’s judgment remained in the loop.

This is exactly what assistant pattern provides. You specify intent; the AI produces an artifact; you review before deploying. The methodology is exposed. The output is inspectable.

The DIY model lacks this verification layer. When you are both the specifier and the executor, there’s no review step. You type the letter yourself. You build the spreadsheet yourself. You create the presentation yourself. If something’s wrong, you catch it—or you don’t.

Executable Knowledge Architecture describes the pattern: “The code is the boundary itself. Inspectable. Testable. Deterministic.” The secretary’s typed letter was that boundary. The AI’s generated artifact is that boundary. The DIY professional working directly in Word has no boundary—just themselves.

The Transition

The cultural resistance will echo the 1980s.

“Real executives don’t type” gave way to everyone typing. “Real professionals do their own work” will give way to professionals who specify, verify, and delegate execution.

The transition costs are significant but not unprecedented. Organizations that moved from secretarial pools to PCs faced training costs, infrastructure investments, and cultural adjustment. Organizations moving from DIY tools to AI assistants will face similar challenges: new workflows, new skills (specification and verification rather than execution), new infrastructure.

The professionals who thrive will be those who recognize the shift: from execution to specification, from production to verification, from doing to directing. The expertise that commands a premium isn’t formatting documents or building spreadsheets. It’s knowing what should be said and confirming that it’s said correctly.

The Choice

For forty years, we’ve accepted the hidden tax of DIY knowledge work. We’ve treated productivity software as liberation rather than burden-shifting. We’ve spent our days wrestling with tools instead of focusing on the work that actually requires our expertise.

The AI assistant offers a different path. Not back to the 1950s—the social dynamics of that era aren’t coming back, and shouldn’t. But back to a division of labor that recognized a simple truth: execution is a specialized skill, and forcing everyone to do their own execution is wasteful.

The question from Making AI Make Sense applies: the direction is clear; the question is whether you’re encoding or being encoded.

This time, the assistant is better than any of us at execution, available on demand, and costs less than the software it replaces.

The professionals who recognize this will reclaim their time. Those who don’t will keep fighting with PowerPoint while their competitors focus on the work that matters.