What AI means for the worker

Most of us work as specialists, inside teams that tell us what to do next and check what we have done. AI lets one person do work that used to need several, which is the work of a generalist, and a small number who learn it early could put many others out of work before they have had time to adapt.

How most of us work today

Most people at work today are specialists. We train for one kind of work, often for years, and we are employed to do that work inside an organisation arranged so that we need to do little else. A payroll administrator is expected to run the payroll accurately and on time, and is not expected to draft the employment contracts or maintain the payroll software, because each of those belongs to someone else whose training is in that field.

The organisation around the specialist does a great deal of work that is easy to overlook. A manager decides what matters this week and in what order. Colleagues with more experience answer questions and notice mistakes before they reach a customer. Procedures record how a task has been done before, and a profession or trade sets the standard for what counts as done properly. Work usually arrives already divided into pieces of a manageable size, and when a piece falls outside a person’s remit there is somebody to pass it to.

Much the same is true of people who work for themselves. A self-employed electrician relies on an accountant for the books and on the regulations of the trade to define a safe installation.

This arrangement is how complicated things get done reliably. It allows each person to become very good at one part of the work without having to understand the whole. It also means that the knowledge of what to do next, and of whether the work has been done well, is held largely by the structure around the worker and only partly by the worker.

We have worked like this for a very long time. Specialisation began with farming, and we have not operated as generalists since we were hunter-gatherers.

What AI lets one person do

Used well, AI makes a person competent across a much broader range of work than they could cover alone. Someone with no legal training can produce a usable first draft of a supplier agreement, and someone who has never written software can build a working tool for their own team. They can have that work checked as they go, by asking for it to be tested, compared against a source or reviewed for errors. Processes they used to repeat by hand can be automated by the person who does them, with no specialist or separate department involved.

A single person now has the capability which, until recently, belonged to a small team, and for anyone running a business or building a career that is a huge opportunity.

How the work of a generalist differs

Using that capability is a different kind of work from the specialised work described above, and the difference is larger than it first appears.

A specialist is handed work that somebody else has already defined and divided. A generalist starts with a problem and has to decide what the tasks are, which of them matter and in what order to take them. That decision normally belongs to a manager, and most people have never been asked to make it for work outside their own field.

A generalist also directs work they could not do themselves. With AI this means asking for a contract, a set of accounts or a piece of software without the training that would tell you whether the result is right. The tools produce work that reads well in every field, and work that reads well is not always correct. The person directing it has to know enough about each field to ask for the right thing, to recognise an answer that is wrong, and to find a way of testing an answer they cannot judge by reading it.

In a team a second person checks the work, and a professional standard says what good work looks like. A generalist working alone has to set that standard for each task and arrange the checking personally, which takes discipline when the work already looks finished.

The parts also have to fit together. In an organisation each specialist delivers one part and somebody else is responsible for assembling them, whereas the generalist answers for the whole result, including the places where one part meets another.

A generalist has to know where their own competence ends as well. There are tasks where a wrong answer is expensive enough that a qualified person is needed, and recognising those tasks is a skill that specialised work rarely teaches, because a specialist’s boundaries are drawn for them.

Deciding what to do, directing work outside your own training, checking it and knowing when to stop are the parts of work that specialised roles take away from the individual and give to the organisation. Most people have therefore had very little practice at them, however good they are at their own job.

Why this is hard to learn

The AI generalist of the twenty-first century is, in that sense, the modern hunter-gatherer, expected to turn a hand to whatever the work requires. Even the hunter-gatherer learned each skill by watching others in the group who already had it. The worker today has no working model to observe first, because very few people have worked this way for long enough to be copied.

At its core, then, the worker’s problem is only partly one of learning to use a new tool. They also have to learn a new and fundamentally different way of working, one that asks for the judgement described above, and they have to learn it without an example in front of them, usually while still doing the job they already have.

A small number will adapt first

Across society there will always be some who excel in emergent domains. They tend to be people who are comfortable working without instruction, and they find their own way into a new technology while most of us are waiting to be shown. That group is often the initial spark when a new technology appears, and the change then spreads from them through wider society over a period of years. In previous cases the early few still needed everyone else, because they could not do everyone else’s work themselves, and that need gave the rest of us time to catch up.

With AI that may no longer hold. If one capable generalist can produce the output of a small team, only a relatively small number of them are needed to cause significant disruption to a market. A new company made up of a handful of such people could compete on price with an established company many times its size, and the staff of the larger company could be out of work before they have been able to adapt. I suspect this is the AI apocalypse that Silicon Valley’s leaders keep warning us about, with what seems to me only partial sincerity.

The jobs we are promised

There will of course be a need for certain technical staff, and these are the jobs politicians tend to point to when a new data centre is proposed. A site that takes a thousand or more workers to build is typically run afterwards by somewhere between fifty and a few hundred permanent staff, which is around a tenth of the construction workforce and often less. Those permanent jobs are useful to the places that get them, but there are nowhere near enough of them to replace the work that may be lost elsewhere.

Why a company that sells AI is saying this

I am the founder of a company which sells AI solutions, and I am writing about potential harm because now is the time when we can act to prevent later cost. If people learn to use AI fluently now, a different kind of transformation becomes possible, one in which the knowledge of how to get the full benefit of AI spreads too quickly for any small group to hold a monopoly on economically valuable activity. People take that knowledge into their own jobs and businesses, and the gains in efficiency are captured by the jobs and businesses that exist today, before they go out of business.

That matters because a business which adopts the efficiency itself is still trading afterwards. It can put the freed capacity into work it could not previously take on, and its people move into different roles within a company that has survived, whereas a business undercut by a leaner competitor closes and its staff lose their jobs together.

What EGM offers

EGM works on this through each of its three divisions.

EGM Software builds bespoke software for sole traders, small and medium-sized businesses and larger organisations, and the client owns what we build, so that an established business gains the efficiency itself before a competitor does. We also train our own people in this way of working. Trainees on our Software Operator Programme, some of them university students, work through twelve modules built around a worked case, which gives them the working model that most people lack.

EGM Consulting reviews the obligations and systems behind a business and gives a written decision on what should change, including where the right answer is a local or cloud AI system built, through the software division, inside the business and under its own control.

EGM Education and Training provides AI training for schools and businesses, with courses for teaching staff, school leaders and corporate teams. We use only trainers with demonstrated experience in education or instructional design, and the courses show the work being done before learners are asked to do it themselves.

If you would like your staff, or your school, to be learning this now, please contact us at enquiries@egmfamilyholdings.com.

Written by the people who do the work at EGM. All Insights · Write to the company