Every generation gets one version of this story, and every generation forgets the ending until it happens to them.
There was a time when being a fast, accurate typist was a career. A good typist could turn handwritten notes into clean business correspondence, reproduce documents, correct mistakes and keep an office moving. Speed and accuracy mattered because every correction, copy and formatting change carried a real cost in time.
Then the word processor arrived.
It did not remove the need for letters, reports, proposals or records. Organisations still needed documents, often more of them than before. What changed was the value of the individual tasks involved in producing those documents. Correcting a sentence no longer meant retyping a page. Creating another copy no longer required carbon paper. Reformatting a report did not mean starting again.
The people who adapted did more than transfer their typing skills to a new keyboard. They learned formatting, templates, mail merge and document management. Some moved into spreadsheets, databases, desktop publishing and broader administrative roles. The technology allowed them to take responsibility for more of the finished result.
Others were left exposed. If someone defined their value narrowly as “I operate a typewriter”, there was little left to protect once the typewriter disappeared. The underlying need had not vanished. Businesses still needed documents. They simply no longer needed the same number of people performing the same narrow process.
The value moves before the job title disappears
AI does not have to eliminate an entire profession to change its economics. It only has to make a valuable task faster, cheaper or easier to reproduce. Once that happens, the market becomes less willing to pay a premium for the task on its own.
When people ask whether AI will take their jobs, they often imagine a direct contest between one human and one machine.
The more immediate contest may be between two people with similar experience, one of whom has learned to use AI effectively.
One employee spends three hours producing a first draft. Another produces a workable draft in 30 minutes, then uses the remaining time to verify it, improve it and consider the consequences.
One analyst reviews a set of documents manually. Another uses AI to identify themes and exceptions, then applies human judgement to the areas that matter.
One small business responds to routine enquiries individually. Another creates a controlled workflow that drafts responses, routes unusual cases and records what happened.
The second person is not necessarily more intelligent or more experienced. They are working with leverage.
This does not mean speed is the only measure of value. A faster wrong answer is still wrong. A polished AI-generated report can contain invented facts, weak reasoning or confidential information that should never have entered the system. The advantage belongs to the person who can combine speed with judgement, verification and accountability.
The scarce skill is moving. It is no longer simply the ability to produce a first draft. It is the ability to define the problem, direct the tool, recognise weak output and turn a draft into a reliable result.
The pattern repeats
The typewriter is not the only place this has played out.
Calculators did not eliminate accountants. They reduced the value of doing arithmetic by hand. As calculators absorbed more of the mechanical work, accountants could spend more time on analysis, forecasting, controls and advice. The value moved from performing the calculation to understanding what the numbers meant.
Spreadsheets told the same story more loudly. VisiCalc, Lotus 1-2-3 and Excel did not remove the need to understand income, expenditure, cash flow or financial records. They made recalculation, comparison and modelling dramatically easier. Manual ledgers became less important. Interpretation, assurance, real-time reporting and scenario planning became more accessible.
The bookkeeper who saw the ledger as the job was vulnerable. The bookkeeper who saw the job as maintaining accurate records and helping a business understand its finances had more room to move.
Photography provides a less comfortable example. Digital cameras did not eliminate photography, but they severely reduced the need for film-processing and darkroom roles. Photographers who adopted digital workflows could review images immediately, edit them, deliver them quickly and expand into video and online content. Some people moved up the value chain. Some supporting jobs genuinely disappeared.
That qualification matters. Technological change is not painless, and adaptation is not a magic shield. Some roles will shrink. Some will disappear. Some people will be displaced even when they are capable and willing to learn.
The historical lesson is not that technology never takes jobs. It is that technology often absorbs specific tasks before the full effect becomes visible in job titles. The human who can move from operating the old process to solving the wider problem has more room to move with it.
Across each of these changes, the tool absorbed a narrow technical skill while the person who adapted absorbed a broader role.
AI really is different
There is a tempting response to every concern about AI: “People worried about calculators and computers too, and we adapted.”
That is not a sufficient answer.
AI is more general-purpose than a calculator, a spreadsheet or a digital camera. It can work across writing, research, software development, customer service, design, analysis and administration. The disruption is not confined to one profession learning one new tool. Many professions are encountering their own version of the same challenge at once.
The pace is also different. The move from typewriters to word processors unfolded across offices over many years. AI capabilities can change between one organisation’s annual budget cycles. A task that seemed beyond automation last year may be routine inside a widely available product this year.
The opportunity to adapt is not equally distributed either.
Someone working two jobs does not necessarily have free evenings to experiment with AI. A junior employee may be told to become “AI-ready” without being given access to approved tools or any useful training. Someone without a suitable computer is not competing on the same footing as an employee whose organisation provides the technology and protected learning time. Workers in regulated settings may also face legitimate restrictions that people in less sensitive roles do not.
The International Labour Organization estimates that one in four jobs worldwide has some exposure to generative AI, while concluding that transformation is more likely than outright replacement. It has also warned that the risks are not evenly shared, including higher exposure among women because of their concentration in some of the most affected occupations.[1][2]
So “adapt or be left behind” becomes an empty slogan if employers provide neither the time nor the tools for adaptation. Organisations cannot demand AI-enabled productivity while leaving staff to learn through trial and error, use unapproved consumer services or take personal responsibility for every risk.
Refusal is not always an individual failing. An organisation can refuse to adapt by buying licences without redesigning any work, blocking useful experimentation, withholding training or treating AI skills as personal homework to be completed outside working hours.
Adaptation is a shared responsibility. Individuals need to remain curious about how their work is changing. Employers need to provide safe tools, relevant training, clear policies and time to learn.
The unequal access to reskilling does not make adaptation less important. It makes early, practical access more urgent. The same speed that makes this transition harder also allows the gap between early movers and late adopters to open much faster.
What embracing AI actually looks like
Embracing AI does not mean using it for everything. It does not mean becoming a software engineer, accepting every answer the machine produces or chasing every new product.
It means learning where the technology genuinely improves your work and where human judgement must remain in control.
1. Define your job by the problem, not the current task
The typist who believed the job was “operating a typewriter” had fewer options than the person who understood the job as “producing accurate, useful documents”.
Ask what outcome people actually rely on you to deliver. Is it entering data, or maintaining an accurate record? Is it writing reports, or helping someone make a sound decision? Is it answering emails, or making sure customers receive a timely and correct response?
Tasks change faster than underlying needs.
2. Start with one real, repeatable task
Do not begin with a vague instruction to “learn AI”. Choose a low-risk task that occurs regularly, such as summarising meeting notes, drafting a standard update, comparing two policies or categorising enquiries.
Measure whether the tool saves time or improves quality. If it does neither, it is theatre rather than transformation.
3. Learn to direct, check and improve the output
Typing a request into a chatbot is not the durable skill. The durable skill is giving the system enough context, setting clear constraints, checking the answer against reliable evidence and knowing when the result should not be used.
The moat that remains once basic production becomes cheap is judgement: knowing when an answer is right, when it is subtly wrong and when it should not be trusted at all.
Judgement and verification are not optional finishing touches. They are the work.
4. Deepen the knowledge the tool cannot own
AI can produce a plausible answer without understanding the organisation, the client, the politics, the history or the consequences of getting it wrong.
Domain knowledge, professional judgement, relationships and accountability become more important when first drafts become cheap. The person who knows what a good answer looks like remains more valuable than the person who can merely produce more words.
5. Redesign the workflow, not just the first draft
Buying an AI licence is not adaptation.
The larger gains come from examining the whole process. What information enters it? Who approves the result? Where is it stored? What happens when the case is unusual? How is confidential data protected? How can someone review what the system did?
An AI tool placed inside a poor process may simply help the organisation make mistakes faster.
6. Use AI safely
Employees should not be forced to choose between productivity and confidentiality. Organisations need clear rules about approved tools, personal data, client information, intellectual property, record keeping and human review.
Where this shows up in Microsoft 365
We see a version of this same pattern every week in the work Adjona does.
The organisations most exposed to risk when they adopt Copilot or automate a workflow are rarely the ones actually using the new tool. They are the ones who never went back to check what access, sharing and permissions looked like before they turned it on. Copilot does not create a new risk out of nothing. It surfaces whatever was already sitting there, unreviewed, and makes it far easier to find.
The pattern is the same as the typewriter, just compressed into months instead of decades. The organisations that treat AI adoption as “buy the licence and switch it on” are the ones most likely to discover a problem after the fact. The organisations that treat it as a chance to review who has access to what, tidy up the parts of Microsoft 365 that have drifted for years, and put proper controls in place before switching anything new on, are the ones who get the benefit without the exposure.
That review is exactly the work Adjona does. Not blocking AI, not slowing anyone down with red tape, but making sure the foundations are solid enough that adopting these tools is a genuine gain rather than a hidden risk.
What to do next
None of this guarantees comfort or job security. No honest argument can promise that every role will survive. It does, however, improve the odds that a person’s skills will transfer when the task, tool or job title changes.
The word processor did not end document work. It made basic document production easier and raised expectations of what one person could deliver. AI is doing something similar across a much wider range of work. Drafting, summarising, classifying and reformatting are becoming less scarce. Framing the right problem, applying context, testing the answer, managing risk and taking responsibility for the result are becoming more important.
The safest position is not to compete with AI at the narrow task it performs cheaply. It is to use the tool while strengthening the human capabilities around it.
The typewriter did not warn anyone before it became a museum piece. It did not need to. The people who paid attention to what was happening around them warned themselves and adjusted in time.
The lesson is not that everyone must chase every new technology. It is simpler than that: do not confuse the machine you operate with the value you provide.