I started testing AI writing tools because I kept ending up with half-written books and no realistic path to finishing them.
At first, I did what most people do. I used general assistants like Claude to draft content. And to be clear, Claude is excellent at writing. The prose is thoughtful, it handles long context well, and it’s great for brainstorming and research.
But after a few attempts, the same pattern kept repeating.
I could generate good text, but I couldn’t turn that text into a finished book without a ridiculous amount of manual work. Every chapter lived in a chat window. Outlines were separate. Formatting happened later in Word or Google Docs. Covers meant opening Canva. EPUB meant another tool. Audiobooks meant yet another one.
None of that sounds hard in isolation. The problem is that when you stack all of it together, momentum dies.
What I realised pretty quickly is that most AI tools are optimised for writing, not for publishing. They stop at text. Everything after that is assumed to be someone else’s problem.
That’s fine if you’re writing an article or a one-off document. It completely breaks down when you’re trying to produce a real book.
The biggest issue isn’t quality. It’s continuity.
Once you’re a few chapters in, you start dealing with small but compounding problems. Chapters drift slightly in tone. Ideas repeat. You’re never quite sure which version of a paragraph is the latest one. You start hesitating before generating the next chapter because you know you’ll have to clean things up later.
That hesitation is what kills most projects.
What finally worked for me wasn’t better prompts or switching models. It was changing the workflow so finishing became the default outcome instead of something you had to push yourself toward.
That meant starting with structure, not text. A proper chapter outline that stays fixed while the content fills in around it. It meant writing chapters in a place where they’re already formatted, already organised, and already part of the final book.
It also meant treating exports as part of the process, not a chore at the end. When you can click a button and get a clean PDF or EPUB at any point, the book starts to feel real much earlier. That changes how you work on it.
This is basically why I built Inkfluence.
Not because Claude writes badly. It doesn’t. And not because tools like Canva or Designrr are bad either. They’re good at what they’re designed for. Canva expects you to bring finished content. Designrr expects you to already have blogs or podcasts to repurpose.
Inkfluence exists for the gap between those tools. When you have an idea, but not a manuscript. When writing is the bottleneck, not design. When you want to go from topic to a professional, downloadable book without stitching together five different apps.
Under the hood, Inkfluence is just using strong language models, the same class of models people already trust for writing. The difference is everything wrapped around them. Chapter management instead of chat messages. A real editor instead of copy-paste. Covers, exports, and audiobooks as part of the same workflow instead of afterthoughts.
The most important thing I’ve learned from all of this is that finishing books isn’t about brilliance. It’s about removing the points where people naturally stop.
If a tool reduces the number of decisions you have to reopen, reduces the amount of manual cleanup, and makes progress feel visible, you’re far more likely to get to the end.
Once you finish one book, the next one is easier. And the next one after that even more so.
That’s the real leverage. Not better sentences, but a system that doesn’t collapse halfway through.
If you’re curious, Inkfluence is the tool I ended up building to solve exactly this problem:
A workflow designed to get books finished.
Top comments (0)