Most writing about AI and small business is either breathless hype or vague reassurance. Neither is very useful when you're actually the one running the operation. Here's a more literal accounting of what changed.
The work that actually moved
Running a software portfolio and an agency at the same time means a constant stream of small, specific writing tasks: App Store listing copy, support email replies, documentation pages, internal specs for whoever's building the next feature. None of this is creative writing in any meaningful sense. It's precise, repetitive, and easy to get subtly wrong.
This is where AI tools have made a real, measurable difference, not by being clever, but by removing the blank-page problem entirely. A first draft of a privacy policy, a set of test instructions for an app reviewer, a redo policy explanation grounded in the actual regulation it's based on, all of these go from a 45-minute task to a 10-minute editing pass.
The work that didn't move at all
Judgment calls didn't get any faster. Deciding whether a new feature is worth building, whether a price increase is justified, whether a vendor's data can actually be trusted, none of that compresses, because the bottleneck was never typing speed. It was always full information and a clear head, and AI doesn't supply either one on its own.
If anything, the danger runs the other way. A tool that produces confident, fluent text on demand makes it easier to mistake fluency for correctness, especially under time pressure. The discipline that actually matters is still asking "is this true" before "does this sound right," and that discipline doesn't come from the tool.
Where it actually saves a solo operator time
- First drafts of repetitive, structured writing: policies, listings, documentation
- Restructuring and condensing existing material into a new format
- Catching small, mechanical errors before they ship
- Modeling out a calculation or scenario before committing to a decision
Where it doesn't
- Deciding what's actually worth building next
- Negotiating, hiring, or anything that depends on reading a real person
- Anything where being wrong confidently is worse than being slow and right
None of this is a verdict on AI generally. It's just a more specific account than "AI saves time," which is true in the way "exercise is good for you" is true: real, but useless without the actual specifics of what, when, and how much.