消失的秘书案
The Case of the Disappearing Secretary

原始链接: https://rowlandmanthorpe.substack.com/p/the-case-of-the-disappearing-secretary

## 自动化浪潮的循环与工作的未来 本文探讨了自动化的历史背景,认为当前围绕人工智能的焦虑与过去的科技变革,特别是20世纪80、90年代的计算机革命相呼应。虽然人工智能感觉独特,但大规模自动化并非首次出现。我们常常忽略了工业革命之外的先前浪潮,例如对文职工作的剧烈重塑。 在计算机出现之前,秘书是必不可少的——是管理者和信息之间的关键接口。她们处理从听写和打字到归档和日程安排的一切事务,代表了劳动力的一大比例(美国和英国约为18%)。计算机化并没有立即*消除*这些工作,而是从根本上改变了它们,转移了任务和期望。 作者的母亲的经历对此进行了说明:从厌恶文职工作的单调和顺从,到当老板们开始通过个人电脑实现自给自足时感到解脱。这种转变反映了人工智能可能带来的未来——“接口”角色的重新组合,可能使每个人都变成“意外的管理者”,监督人工智能代理。 关键要点不是工作*减少*,而是工作*转型*。正如计算机化创造了新的角色(和“影子工作”),人工智能很可能重塑专业知识要求,可能提高专业角色的工资,同时降低其他角色的工资。历史表明,这种转变将是混乱的、不均衡的,并且最终是长期技术颠覆模式的延续。

Rowland Manthorpe 的 Substack 上一篇名为“消失的秘书”的文章,在 Hacker News 上引发了关于秘书角色演变的讨论。用户 rwmj 分享了这篇文章,促使 ghaff 发表评论,分享了自己的经历。 Ghaff 指出,在其职业生涯中遇到的秘书质量各不相同,甚至在电脑普及之前就是如此。有趣的是,他们还记得高层管理者*要求*将邮件打印出来,手写回复,然后由秘书重新输入用于内部沟通——这种做法说明即使在技术出现后,仍然依赖行政支持。 评论者同意秘书任务已经分散到许多员工中的观点,但质疑共享助理的整体实用性,尤其是在电脑工具不断改进的情况下。这场讨论凸显了行政工作处理方式的转变,以及专门秘书角色可能面临的衰落。
相关文章

原文

Something different this week. This is an expanded version of a talk about AI that I gave recently at Sky Media. After I finished I realised I needed to investigate further, because – well, you’ll see why.

It’s possible that artificial intelligence is something unique in human history, but the mass automation it seems bound to produce definitely isn’t.

Despite this, we rarely hear in any detail about previous waves of automation. There’s discussion of the Industrial Revolution, but that’s about it. We hear more about Engels’ Pause than we do about flagmen or telephone operators or motion picture projectionists.

This seems strange, because there has been a huge wave of automation within living memory. In fact, we are still living through it.

To see what I mean, take a look at this map of the most common job in each US state in 1978.

With the exception of truck drivers – for now – every job on that map has been reshaped by automation. (Globalisation played a role too, but it’s far from the whole story.) There aren’t as many machine operators around any more. Nor farmers. And there definitely aren’t as many secretaries.

Their fate is the subject of this essay, and a lens to think through the implications of AI for work with a bit more nuance than “LLMs are a scam” or “white collar work is doomed.” Perhaps those all-or-nothing predictions will turn out to be right! But honestly I doubt it. Instead I think it will be messy, confusing, exciting, strange, unfair and apparently irrational, just like it was last time.

Not so long ago, the work of secretaries – typing, filing, organising, administrating – was a cornerstone of the economy. By 1984, six years after the map above, there were around 18 million clerical and secretarial workers in the United States, roughly 18 percent of the entire workforce. This was totally normal. In the UK at the same time, between 17 and 18 percent of the workforce was some kind of secretary. In France it was 16 percent. Different economies with different economic policies; all ended up with one in five or six workers employed in clerical work.

Why so many? Because every stage of information processing required a human hand. In a mid-century organisation, a manager did not “write” a memo. He dictated it. A secretary took it down in shorthand, then retyped it. Then made copies. Then collated the copies by hand. Then distributed them. Then filed them. And so on and so on. Nothing moved unless someone physically moved it. There was no other way.

For this reason, the most sophisticated, information-dense organisations were often the ones with the most administrative staff. As NASA prepared to launch the Apollo missions in the mid-1960s, 15% to 18% of its civil service workforce was classified as “clerical and administrative support”. There were the human “computers” made famous by Hidden Figures, but also technical typists, who typed up mathematical equations. As one of those typists, Estella Gillette, later put it: “The engineers depended on us for everything that wasn’t their job. We were their support system.”

This line is often taken as an inspiring motivational quote, but it was a literal description of the situation at the time, because of what today we might call an interface problem. The invention of shorthand and the typewriter in the early twentieth century had made it possible to create accurate records, but senior staff – even engineers at NASA – didn’t interact directly with the administrative machinery of the office. Secretaries and clerks were the unavoidable interface between the manager and the ability to get things done. You spoke to a secretary; they “interfaced” with the shorthand pad and the typewriter. You handed over a paper; they “interfaced” with the filing cabinet. Every kind of activity was organised this way. The secretary was the interface for the diary, a physical object kept only on their desk. (This could be a source of real influence.) They were the human “firewall” or routing system for phone calls. If the manager wanted a coffee, well that was the secretary too. It all went through her.

Then came the personal computer.

At first the shift to PCs must have seemed almost laughably crude, as physical filing cabinets were duplicated on primitive un-networked computers. But bit by bit the computer and its offspring the internet automated administrative tasks, until eventually many were obsolete.

To find out what this felt like, I asked someone who worked as a secretary during that era: my mum. When she left school in 1972, her parents advised her to seek steady employment, so she attended secretarial college to learn typing and shorthand. She hated it. Then she became a secretary and she hated that too. It wasn’t just the relentless sexual harassment – ”oh yes, that was the norm” – it was the mind-numbing deference and boredom. “You typed a letter, then you put it in a blotter book for your boss to sign, he signed it, then gave it back to you…. One of the worst things was being called in for dictation by someone with a total inability to string a sentence together… It was life sapping.”

For the first decade of my mum’s working life nothing much changed. Then she went on maternity leave in 1982 and, when she came back to work, everything was different. The bosses had started doing their own typing, “seemingly overnight”. To us this might seem like a small thing, but in this world it was everything. The feudal system of the secretarial age – ”secretary gave status to boss, boss’s status reflected on her, typing pool gave nothing,” my mum recalled – was about to disappear forever.

Now, the interface with the machinery of work is changing once again: from the computer to AI. This isn’t meant as a grandiose statement about the all-encompassing power of AI. I mean, simply, that if you want to get things done, it’s increasingly obvious that the best way is going to be through some kind of conversation with a machine, especially when the machine can then go and complete the task itself. Think of an admin-enabling app, whether it’s Outlook, Teams or Expedia. It’s hard to see a future where they’re not either replaced or mediated by AI.

The computer era unbundled the interface known as “the secretary”. The next era may rebundle it back into AI.

These are the lessons from the last change for the new one.

Think we’re the first generation to dream of a workless world? Not at all. “The constant mantra was the wonder of the paperless office and everyone would have more leisure time,” my mum recalled. A 1986 National Academies of Sciences, Engineering, and Medicine paper on new workplace technologies reported widespread claims that “in the foreseeable future, productivity may be so enhanced that employment may become a rarity for everyone.”

Reader, it was not.

Almost two million non-legal and medical secretaries in the US alone. And not just secretaries - administrators, executive assistants, clerks of different kinds, as well as typists and word processors.

Yes: according to the Bureau of Labor Statistics, there are still around 45,000 people in the United States whose primary occupation is typist or word processor. That’s only 0.025 percent of the workforce, down from 250,000 at the turn of the millennium, but still – they exist. Technological displacement takes a long time to produce literal extinction. An obvious point, but an important one.

While sellers of machines like word processors hyped up the potential boost to productivity – up to 150 percent increase in secretarial output! – most sensible observers saw little prospect of deep and lasting change for secretaries from computerisation. “The variety of the tasks and the social relations on the job have led to little labor displacement, and little is likely in the future,” concluded the National Academies report, comparing secretaries to nurses in their indispensability.

Tellingly, “secretary” isn’t a standalone category any more in headline UK labour statistics, which makes it difficult to work out exactly how many secretaries of the classic type there are; and in any case the job has changed so much it’s hard to make comparisons at all. But according to the 2021 census for England and Wales, 238,210 people were classified as personal assistants, secretaries or typists, roughly 0.9% of the workforce. In the US the Bureau of Labor found 1,785,430 secretaries and administrative assistants in 2023, around 1.1% of the workforce.

The job my mum did still exists, but perhaps not for much longer.

Although officially secretaries are now only around one percent of the employed workers, it wouldn’t be right to say that the administrative share of the workforce has gone from one in five or six to one in a hundred. Not at all! There are still lots of administrators around; they just have different names. Human Resources. Business operations. Compliance. Routine clerical labour became “professionalised administrative coordination.” (Although it was still done mostly by women. Economic change does not always imply social change.)

This is the classic pattern of automation, seen everywhere from farming to the military. You stop doing tasks and start overseeing systems.

Will the same thing happen with AI? If you look at software engineering, it’s clear it already is.

Most secretarial work wasn’t removed; it was spread around so that everyone did it. If you work in an office today (and even if you don’t), you do your own typing, your own formatting, you send your own emails, you arrange your own meetings and you answer your own phone calls. If you go on a work trip, you probably book your own flights, your own accommodation and when you’re back you file your own receipts.

The scale of this “shadow work” is immense. Imagine travelling back in time to explain that, over a stiff gin and tonic, to a mid-level manager in the 1970s. They would look at you like you’re mad. “You’re telling me this and you say things have got better??” And that’s even before we get to the work created by computers - the endless emails, the meetings which should have been emails, the emails to arrange the meetings which should have been emails, and so on.

So what will be the shadow work of the AI era? An obvious candidate: management. Boris Cherny, who leads Claude Code, doesn’t code anymore. Nor do lots of people at Anthropic. So what do they do? They manage their non-human teams.

A lot of engineers talk in exalted terms about the feeling of power this gives them. I’ve heard the phrase: “it’s like being the conductor of an orchestra.” I wonder if it will still feel that way when the novelty wears off and the work of supervising and dealing with agents is just another branch of working life. Professor Ethan Mollick calls management an “AI superpower”, but it seems to me that you might also call it an AI chore, something we will have to do even if we don’t want to, that’s by turns draining, frustrating and stressful, and creates as much work as it is supposed to eliminate. As the authors of a recent study put it: “AI Doesn’t Reduce Work—It Intensifies It”.

Computerisation turned everyone into an accidental secretary. AI will turn everyone into an accidental manager.

Regardless, it seems that this is the way things are heading. Computerisation turned everyone into an accidental secretary. AI will turn everyone into an accidental manager.

The personal computer did not immediately reduce administrative employment, it increased it. Some groups of administrative workers – stenographers, for instance – went into terminal decline, but as the economy boomed in the 1990s, the demand for administrative coordination actually went up, a Jevons Paradox for bureaucracy.

It wouldn’t surprise me if we saw something similar for software with AI; indeed job postings for software engineers are already rising in both the US and UK. Of course even in this optimistic scenario, there will still be a lot of fear and dislocation, just as there was in the 1980s and 1990s. Many secretaries were put out of work and many managers found the loss of their “office wife” painful (“If there is anything a man hates, it is to give up his secretary,” said Evelyn Berezin, the builder of the first computerised word processor). Still, the shock was cushioned because there were opportunities for those that went with the change. It wasn’t until later that computerisation began shrinking the broader administrative workforce, because–

Think of the phrase, “on the same page”. Like a lot of sayings – “kick the bucket”; “bite the bullet”; “cut and paste” – it was originally a purely literal description, because making sure everyone had the same page was an essential part of the typewriter era. If NASA updated a manual, someone had to find every copy in the building and swap out “Page 42” with a new “Page 42”, or face potentially disastrous consequences.

For a long time, computerisation changed very little. The first word-processers were really just typewriters with screens: the typist could go back and change the text but everything was still printed in the same way it had always been. At length, computers were able to display digital representations of pages, but although these could in theory have taken many forms, for a long time nothing much changed. Even today there are still plenty of Word documents attached to emails and pdfs with names like, “version 4 final FINAL do not touch”. (Many government press releases take that form.) There are pages and it takes effort to keep them current.

Increasingly, however, the phrase “on the same page” is becoming as divorced from its origin as “hang up the phone”. We are shifting away from pages towards chats and threads; even where we do have pages, they are often stored on cloud systems which make the very idea of out-of-sync copies structurally impossible. (Those systems also automatically scan every word in a document and make them searchable, thereby eliminating the entire task of filing and document retrieval.) The work of staying literally on the same page is being gradually made obsolete.

This shift took decades. Yet although generative AI is, by many measures, the fastest technology ever adopted, that doesn’t mean it will skip the awkward in-between stage. Will AI eventually displace all software in some form? Perhaps – but right now Anthropic and OpenAI use Workday for their HR, so I think it’ll survive a while yet. Are those websites that have a chatbot ready to help (or, just as often, hinder) the final form of this interface? Probably not, but if history is any guide we might be stuck with them for some time.

As computerisation kicked in, secretaries found themselves being asked to take on responsibility for work in larger teams. Some welcomed this change. “Younger secretaries tend not to like the subservient role and are far less likely to be prepared to put up with the behaviour of some very difficult senior executives,” the British Institute for Employment Studies found in 1996. Others found it difficult, as a 1994 Guardian article reported:

Secretaries used to be part of the office furniture, seen but rarely heard. . . . A good secretary was an unremarkable one, efficiently obeying orders, and then returning mouse-like to her station behind the typewriter. . . . Now they [secretaries] are becoming a key part of the team . . . With lots of people competing for a secretary’s time, he or she will need to exercise assertiveness and understand the dynamics of organising the workload of a group

Being moved – or pushed – into a coordination role was better than the alternative. During the first wave of computerisation, many secretaries found that the new technology chained them to their screens, turning the office into an “assembly line”. What’s more, the new computers allowed managers to watch secretaries more closely. From a Washington Post article with the headline “Computers Said To Zap Clerical Jobs”:

Office workers nowadays are doing more work with their new machines. But that productivity usually encourages managers to add more assignments in the belief that the machines and the people using them are capable of handling the load. To ensure that the extra work is done, some companies are using computers to monitor the people using the computers.

It might read like it was written yesterday, but this article was from 1986.

Computerisation brought a shift in standards. “While IT has reduced the amount of typing secretaries do,” the 1996 report observed, “expectations about the quality and accuracy of the work produced have increased considerably.” A universal truth: the more capacity we have, the higher our expectations are.

One practice which faded as the typewriter era drew to a close: detailed minute-taking. When every manager had a secretary, it made sense to ask her to record meetings verbatim using shorthand. When they didn’t, this task became seen as an inefficient use of time. “In some ‘action’ meetings a few ‘flagged-up’ bullet points are seen as sufficient record, and these are often taken down by managers,” the Institute for Employment Studies noted in a tone of some surprise.

Funny to think that AI is bringing back the minuted meeting, only this time in the form of transcription. This simple change alone has the potential to spawn a whole industry and a whole new way of working which is invisible to us at present.

Here’s a puzzle. As computerisation hit, accounting clerks and inventory clerks in the United States were both equally exposed to automation. Yet between 1980 and 2018, accounting clerks saw rising wages, while inventory clerks saw their wages fall. How can the same effect produce different results?

The answer, according to economists David Autor and Neil Thompson, depends on which parts of a job get automated. If the highest-skilled aspects of a job are handed over to a machine, then the threshold for entering it falls, allowing people to come in more easily. The supply of labour rises and wages fall. If the lowest-skilled aspects are automated, then the entry-level jobs are the ones that disappear. The industry becomes harder to enter, the supply of labour falls and wages rise.

This was what happened in the case of the clerks. Inventory clerks saw higher-expertise tasks like working out the price of goods displaced by automation, leaving behind mostly generic physical tasks – that’s why their wages fell. Accounting clerks, by contrast, found that computerisation mostly automated routine tasks like data entry and basic bookkeeping, leaving behind tasks which needed more specialised problem-solving and judgement. Their wages increased while their employment declined.

Early evidence suggests that this same dynamic is playing out again with AI. A recent paper by Bouke Klein Teeselink and Daniel Carey using data on hundreds of millions of job postings from 39 countries found that “occupations where automation raises expertise requirements see higher advertised salaries, whereas those where automation lowers expertise do not.”

Of course you’re wondering which jobs will be hit in which way, and Klein Teeselink and Carey do give some examples. This is ChatGPT’s version of their chart. (I write every word by hand but I need help for the charts.) In short: among those with high AI exposure, they expect wages to rise for human resources specialists and fall for – yes – executive secretaries. The wheel turns once again

So, in summary: computerisation ended some jobs, changed lots of others and created many ones. Yet that description covers so little of what really happened, because the biggest change wasn’t to the jobs, it was to the people and how they behaved. This is what I really learned writing this piece. I went in expecting to find out about tasks and technologies and I came out having learnt about a strange world very different from my own, a world now almost entirely vanished.

Accounts from that time, including my mum’s, emphasise that side of things much more than the dry economic account. One oral history from a secretary called Cynthia who worked from 1958 to 2005 mentions how, once, people used to knock at the door of the office – of course the manager had a separate office – and wait to be called. Then, suddenly, they started walking in because they wanted to speak to him directly. That is the world that computerisation helped to bring to an end, and now it is almost impossible to imagine it existed.

The last word has to go to my mum. What happened to her after the bosses started typing? By chance, she was working for a company which leased computers to businesses. She moved into sales and, as computerisation boomed, she escaped the world of the secretary, to her great and lasting relief. She ended up being successful in several other occupations – but that is another story.

联系我们 contact @ memedata.com