Using AI /
Small teams of high-context generalists
For a long time the way to build a team was to add a specialist for each function. One for product, one for engineering, one for design, one for marketing, one for support. Each seat held a job nobody else could do.
AI is quietly loosening that rule. Not by making specialists worthless, but by making one capable person slightly less bad at the jobs on either side of their own.
The old assumption
The standard shape of a company assumes work is divided by expertise. A task arrives, it gets routed to the person who owns that function, and it waits in their queue. The handoffs between functions are where most of the time goes.
That shape made sense when producing a first draft of anything took real skill. You could not write the marketing copy without a marketer, or stand up the landing page without an engineer. The specialist was the only door into the work.
What changes
AI does not make everyone good at everything. It makes a capable person less bad at the neighbouring jobs, and that is often enough to move a project forward.
- The engineer can get a usable first draft of the launch email.
- The marketer can stand up a rough version of the page without waiting in a queue.
- The operator can pull the report themselves instead of filing a request.
None of these drafts are as good as the specialist would produce. They do not need to be. They need to be good enough to keep the work moving, and to make the eventual specialist review faster.
The high-context generalist
The person who gains the most from this is not the one who knows the most tools. It is the one who holds the most context.
A high-context generalist is:
- Technical enough to build, or to direct a model that builds.
- Commercial enough to choose the right problem in the first place.
- Careful enough to know when a decision needs a real specialist.
The value is not breadth for its own sake. It is the judgement to steer work across functions without losing the plot. AI supplies the first draft in each domain. The generalist supplies the taste to know which draft is wrong, and the context to know why it matters.
This rewards people who understand the work
It is tempting to read this as "learn the tools." That is the shallow version.
The tools change every few months. What does not change is understanding how work actually flows through a business: where the real constraint sits, which decisions are reversible, what a good outcome looks like. AI can produce a competent draft in a field you half-understand. It cannot tell you whether that field is even the right place to be spending the effort.
That judgement comes from having done the work, not from having read about it.
The trap: fake generalism
There is a failure mode here, and it is easy to fall into.
Fake generalism is shallow confidence across too many domains. The model hands you a plausible-looking answer in law, in security, in finance, and you ship it because it reads well. The high-context generalist knows the difference between a domain they can carry and a domain where a wrong call is expensive or unsafe.
- Legal, safety, security and high-risk decisions still need a specialist's review.
- Knowing where that line sits is part of the skill, not an admission of weakness.
A generalist who cannot name the edges of their own competence is not a generalist. They are a liability with good output.
The rule
Do not try to become the person who knows every tool. Try to become the person who holds enough context to steer the work, and enough judgement to know where your competence ends.
AI rewards range, but only range with a centre of gravity. The valuable worker on a small team is the high-context generalist: capable across functions, honest about the edges, and able to tell a good draft from a dangerous one.