Remote Work Was Already News in 1979

People Were Already Talking About Remote Work in 1979

On December 11, 1979, the BBC current-affairs program Nationwide covered how word processors and telecommunications networks were transforming the office. The footage, which the BBC has since made public in its archive, shows document processing at Bradford City Council and a programmer working from home.

The segment introduces the possibility that cutting paper documents and exchanging data over phone lines could let people handle office work without commuting.

In 2026, I keep hearing that AI agents can handle code, reports, and email. The tools and the tasks they can do have changed, but the expectation that new technology will cut down working hours feels familiar. Comparing these two moments made me think about how time saved on individual tasks actually translates into organizational performance.

To see that connection, you need to look not just at how well the technology performs, but at how work gets distributed, how review and approval processes function, and how performance gets measured.

All checks pass — the fragment is accurate, complete, and glossary-compliant. No edits needed.

Early Examples of Word Processing and Remote Work

The footage introduces the impact of the word processor system that Bradford City Council introduced in 1977. It explains that the system cut the staff needed for the task and raised productivity by 40%. The method was to turn repetitive correspondence into standard templates and rewrite stored content as needed. This shouldn’t be read broadly as meaning the council’s entire workforce was cut in half.

For a remote-work example, the footage features F International. It shows more than 600 freelancers programming from home, writing code and transmitting it over phone lines. What feels routine now—remote collaboration—was already being attempted back then.

France’s plan to distribute home information terminals belongs to the same wave of change. It was a project to replace paper phone directories with an electronic service. That said, it’s not true that 30 million units had already been distributed by 1979. Later records from the French Parliament show that the actual number of Minitel terminals was still only in the millions even into the 1990s.

The footage also touches on the prospect that working from home could reshape family life. Both hopes—that the workplace would become location-independent—and worries—that technology might homogenize daily life—appear side by side.

The paperless office, remote work, and changes to administrative work were all topics under discussion even then. Still, we need to distinguish between the projections of that era and the changes that actually took hold.

Why Every AI Productivity Study Shows a Different Number

Recent AI surveys keep producing big improvement figures. But what the number actually means depends entirely on what was asked, and of whom.

The study METR published in May surveyed 349 people working in technical fields — developers, researchers, graduate students, and others. The median self-reported change in work speed was 3x. The median change in the value their work produced ranged from 1.4x to 2x, depending on how the question was framed. Both figures are self-assessments by respondents, not measurements of actual hours worked or revenue generated.

Firm-level surveys need the same scrutiny of methodology.

Researchers from Duke University’s Fuqua School of Business and the Federal Reserve Banks of Richmond and Atlanta surveyed roughly 750 corporate executives. The average labor-productivity improvement from AI in 2025, as reported directly by executives, was 1.8%. Calculated instead from executives’ reported effects of AI on revenue and employment, the figure comes out to about 0.6%. Both numbers come from the same survey. This isn’t an experiment comparing perceived impact against externally verified performance. The researchers noted that efficiency or quality gains may simply take longer to show up in revenue.

PwC surveyed 1,217 senior executives across 25 industries. Analyzing responses on revenue and efficiency gains, the firm reported that 74% of AI’s economic returns were concentrated in the top 20% of companies. This shows a gap in performance between firms. It does not mean the remaining 80% were stuck in pilot mode or saw no benefit at all.

The University of Pennsylvania’s Penn Wharton Budget Model offers a different kind of number altogether. Its 2025 model estimates that, compared to a world without generative AI, U.S. productivity and GDP levels in 2035 would be about 1.5% higher. That’s not a 1.5-percentage-point boost to annual growth — it’s a long-term projection built on specific assumptions.

Individual work speed, firm-level labor productivity, and national GDP are different things measured in different units. Lining these numbers up side by side doesn’t let us conclude that AI’s effects simply evaporate somewhere inside organizations. What it does confirm is that before anyone talks about “results,” they need to decide what, exactly, they’re measuring.

Why It Takes Time for Technology Adoption to Turn Into Productivity Gains

In 1987, economist Robert Solow pointed out the gap between the widespread adoption of computers and productivity statistics. This became known as the Solow paradox.

The problem was that buying a lot of computers alone couldn’t explain an increase in economy-wide productivity.

Behind this lie several factors: the speed of technology diffusion, how it’s actually used, and how statistics capture its effects. A slowdown in productivity growth can’t simply be attributed to computer adoption itself.

The conversation shifted once U.S. productivity growth accelerated in the mid-1990s. McKinsey focused on changes within specific sectors like tech, retail, and wholesale. Their explanation: what mattered for performance wasn’t just buying new equipment, but also overhauling processes like inventory management and logistics. Not every industry went through the same lag or reaped the same benefits.

Economic historian Paul David offered a similar explanation for the long lag between adoption and results in factory electrification. Factories that once ran multiple machines off a single steam engine could simply swap in an electric motor as the power source — or they could redesign their entire layout to give each machine its own motor. As the latter approach spread, so did the range of ways electricity could actually be put to use.

Research on the productivity J-curve examines the investments required for this kind of transition — training, workflow redesign, new product development. These investments demand cost and time, and the knowledge and organizational capability they build up may not show up cleanly in statistics. This explains the pattern where measured productivity is low early in adoption and the gains only appear later. It isn’t a law stating that every investment eventually pays off.

Something similar could happen with AI adoption. But it’s hard to conclude from a handful of surveys that we’re currently at the bottom of the J-curve, or to predict that AI will take less time simply because computers took less time than electrification did. The changes required differ from one job and one organization to the next.

Oswarld’s Lens

As I’ve built technology strategy, I’ve come across this kind of performance promise often.

I’ve reviewed plenty of proposals claiming, “Adopting this technology will boost productivity by N%.” But in some cases, it’s not even clear what’s being measured as productivity. Whether it’s document-drafting time, number of items processed, or revenue makes for a completely different story — yet it all gets collapsed into a single figure.

In the cases I’ve examined, the change that actually drove results wasn’t just swapping out tools. Meeting formats, decision-making pathways, and the criteria for checking outcomes all shifted together. Even if AI produces a draft faster, if review and approval still take just as long, total work time may not shrink much at all.

What caught my attention in the Bradford case was the same thing: standardizing repetitive documents and reusing stored content. PwC’s research similarly notes that high-performing companies address how AI is used across both operations and the business as a whole. That said, this shared pattern alone isn’t enough to conclude that organizational redesign explains every gap in performance.

That’s why I try to first decide what needs to change after adopting a given technology. The design looks entirely different depending on whether the goal is cutting processing time, reducing errors, or speeding up customer responses. Optimism that results will eventually materialize if you just wait isn’t enough on its own.

Closing

The 1979 word processor and 2026 AI are different technologies. But the challenge remains the same: connecting the new tool’s performance to organizational results.

Before using terms like AI-Native or Agentic, I think we need to be able to explain concretely what work actually changed and how. Sitting through a few lectures or seminars doesn’t immediately change how work gets done. New technologies will keep emerging, so I’d rather spend my time checking how they’re applied and what results they produce than debating what to call them.

Where has the time your current AI use saved actually gone? If it cut writing time but increased review time, you can start by measuring both together.

Have you tried distinguishing between what merely felt easier with AI and what actually changed your work outcomes? I’d love to hear your experience in the comments.

References & Further Reading

Primary sources

• BBC Archive, “1979: Will Word Processors Start a Home Working Revolution?”, Nationwide, 1979.12.11. : This is the office-automation report that sparked today’s newsletter. You can see period footage of document processing and early telework.
• Baslandze, S. et al., “Artificial Intelligence, Productivity, and the Workforce: Evidence from Corporate Executives”, NBER Working Paper 34984, 2026. : A study analyzing AI adoption and perceived performance based on a survey of roughly 750 executives.
• METR, “Measuring the Self-Reported Impact of Early-2026 AI on Technical Worker Productivity”, 2026. : A survey that separately asked respondents to self-assess work speed versus work value.
• PwC, “2026 AI Performance Study”, 2026. : A global survey of 1,217 executives finding that 74% of AI’s value is concentrated in the top 20% of companies.

Background

• Brynjolfsson, Rock, Syverson, The Productivity J-Curve : Explains how investments in intangible assets — like process redesign and human capability — affect how we measure productivity.

• French National Assembly, Report on the Implementation of Telecommunications Regulation, 1999. : Contains the adoption figures for Minitel, France’s early videotex network.

• McKinsey Global Institute, “Is the Solow Paradox back?”, 2018. : The cleanest available summary of the Solow paradox’s history and structure.

• Fortune, “Why AI is raising worker productivity but not making the economy more efficient”, 2026. : A recent piece analyzing the current 2026 landscape through the Solow paradox framework.

• Penn Wharton Budget Model, “The Projected Impact of Generative AI on Future Productivity Growth”, 2025. : A model estimating AI’s long-term contribution to GDP — the source of the “1.5% by 2035” figure.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.