By Susan Jenkins, 10 September 2025
Intelligent Editing recently released a new version of Draftsmith, the AI-supported editing tool that I reviewed in March 2024. Back then, generative AI tools for working with texts were announced on almost a daily basis, but were still early in the adoption phase by everyday users. Since then, we’ve moved past questioning whether generative AI is going to make much difference in our daily lives. Understanding how to work with it is becoming an essential skill.
Purpose-built tools make adoption easier by reducing the time needed to craft skilful prompts for generative AI chatbots. Remember a time when you bought something you had to assemble yourself, with your own tools, before you could use it for its intended purpose? Did you wish for a complete, ready-to-use item out of the box?
Draftsmith initially launched with a broad selection of ready-to-use tools called ‘suggesters’ aligned to different editing contexts. It took some experimentation to learn which suggesters were suitable for my many wordsmithing hats, but the results were more reliable than when using a Large Language Model (LLM) chatbot. It was also a bit speedier than my unassisted editing process. Draftsmith 2.0 makes the revising process even smoother by improving the flexibility and navigation of the software. They’ve also added a better tutorial on their site.
On the back end, they’ve improved the quality of suggestions by switching to one of OpenAI’s small language models more suited to writing, which also makes Draftsmith’s processing faster.
In my review of the first version, I sampled its arsenal of suggesters to address the various style, audience, or readability aspects that editors pay attention to. For this update, I again tested a few different texts from my workflow: a magazine article, a furniture catalogue, and a research article by a non-native English speaker.
The main feature that’s changed in Draftsmith since version 1.0 is the editing window and the amount of text it processes. Instead of displaying one sentence at a time, it now analyses a whole paragraph. Sentences from the paragraph appear in the ‘decision box’ with suggested changes. Just as in version 1.0, you can toggle the view to see or hide changed text, but there are more options to support your workflow.
Each sentence in the paragraph appears framed between a purple bar on the left and green bar on the right. Purple resets the text to the original, while green generates a new version. You can use these respectively to either reject an edit or regenerate a new one. These functions work on both desktop or tablet devices by using either a mouse, touchscreen, or keyboard to swipe or click for the desired action.
Rejecting an edit with a mouse click (before and after clicking)
Sometimes you like only part of a suggestion, so Draftsmith allows you to tweak the suggestion manually. Simply double-click on a sentence in the decision box and insert the cursor where you want to type, just as you would in the main document.
Once you are happy with all the sentences in the paragraph, click the ‘accept all’ button. Draftsmith will update the document and move to the next paragraph. If you have track changes active, these will be highlighted – another new addition to the software which gives editors more control. By analysing whole paragraphs instead of sentences, you also move through a document more quickly than before.
The new version isn’t without some hiccups. When using the Word Count suggester on quoted text, it sometimes edited the punctuation at the end of sentences unnecessarily. In fact, quoted statements are not always ignored for cuts or word changes, which hampers editing in texts with interview subjects. Draftsmith’s engineers are considering a setting for this in a future version.
Something that hasn’t changed is Intelligent Editing’s customer-centred approach to development. As I wrote in the first review, they are keen to support a very specific group – human editors – and their skills and pain points. While there are hundreds of products on the market for writers and editors claiming AI productivity gains, this company was trusted in the field long before anyone ever thought to connect an LLM with a chatbot.
|
Blog post by: Susan Jenkins |
