Most writing about artificial intelligence is aimed at companies with thousands of employees and a research budget. If you run a company of twenty or two hundred people, almost none of it applies to you. The question is not whether AI will restructure your industry. The question is what it changes on a Tuesday morning in your office.
In our experience, the honest answer is narrower and more useful than the headlines suggest.
It changes who does the first draft
Almost every piece of knowledge work in a company starts with a blank page. An offer letter. A project summary. A reply to a customer complaint. A specification. Producing that first version is slow, and it is slow for a specific reason: starting is harder than editing.
This is where AI genuinely shifts the work. The first draft stops being something a person produces from nothing and becomes something a person corrects. For people who write a lot as part of their job but are not writers by trade, this is a real change in how the day feels.
Note carefully what has not changed. Someone still has to know whether the draft is right. The expertise did not move.
It changes what counts as a small job
Every company has a backlog of things that would be useful but are never worth someone's afternoon. Summarising the last two years of customer complaints by theme. Rewriting the internal handbook so new staff can actually read it. Translating the product documentation into Czech. Checking two hundred supplier contracts for a particular clause.
These tasks were never impossible. They were simply never worth the hours. When the cost of a task drops far enough, work that was permanently postponed becomes work that happens on a Wednesday. In practice, this is where companies find value they did not expect.
It changes where your bottleneck sits
This is the part that surprises people. If you speed up drafting and analysis, you do not automatically get a faster company. You get a company where the bottleneck has moved somewhere else — usually to review, approval or a decision that only one person is allowed to make.
We have seen teams automate a process successfully and gain nothing, because the output now queues in front of a manager who was already overloaded. Before you automate, find out what happens immediately after the step you are speeding up.
What it does not change
It does not change accountability. If an AI-drafted offer contains a wrong price, that is your company's offer with your company's mistake in it. Every serious AI deployment has a review step for exactly this reason, and designing that step well is more important than choosing the model.
It does not change the need for people who understand your business. AI is good at producing plausible work. Judging whether plausible work is correct requires someone who knows why your business does things the way it does.
And it does not change your obligations under data protection law. The moment employee or customer data goes into a system, the question of who processes it, where and under what contract becomes real.
A reasonable first step
Pick one process that is repetitive, heavy on text, and slow today. Write down how it works now, including the handovers. Then ask what would actually improve if the first draft appeared automatically — and who would check it.
If you cannot answer the second half of that question, you are not ready to automate that process yet. That is a useful thing to find out before spending money, not after.
