This is where AI writing tools are at their most useful. Not because they replace your writing, but because they remove the parts that slow you down. The blank page. The fifth attempt at getting the tone right. The nagging feeling that an email sounds harsher than you intended. The report that's basically done but needs tightening up before it goes anywhere.
You've already been using chatbots for writing in Things 2 and 3: drafting content, iterating on tone, shaping output through conversation. This Thing is about a different layer. The tools that work alongside your writing, inside the apps you already use, offering suggestions and rewrites as you type. And it's about getting more deliberate with how you use chatbots for writing tasks, now that you have some experience.
Two kinds of AI writing help
It helps to distinguish between two approaches, because they work differently and they're useful for different things.
Writing assistants: the ones that sit in your editor
These are tools that integrate into the places you already write: your email client, your word processor, your browser. They watch as you type and offer suggestions, or you can highlight a passage and ask them to improve it.
The common thread is that these tools meet you where you're already working. You don't need to copy text into a separate application and paste it back. They're designed to fit into your existing workflow, not interrupt it.
Chatbot-assisted writing: the conversation approach
The other way to use AI for writing is the approach you've already practised: opening a chatbot, describing what you need, and working through it in conversation. This is more flexible than a built-in assistant and often produces better results for substantial pieces of writing, because you can provide much more context and iterate more freely.
The chatbot approach works best when you're starting from scratch, when you need a first draft, an outline, or a set of options to choose from. It's also good for tasks where context matters a lot: a tricky email where the backstory is important, a report that needs to land a particular way with a specific audience, or any piece of writing where tone is as important as content.
In practice, most people who use AI regularly for writing end up using both approaches. The built-in assistant for quick fixes and polish: catching errors, tightening phrasing, adjusting tone on something you've already written. The chatbot for heavier lifting: generating first drafts, brainstorming structures, or working through a piece of writing that needs real thought.
What AI writing tools are actually good at
It's easy to think of AI writing assistance as either a glorified spell-checker or a robot that writes everything for you. Neither is quite right. The most useful way to think about it is as a set of specific capabilities, each of which saves time in a different way.
Changing tone without changing meaning. This is probably the single most useful feature for professional writing. You've written an email that's factually correct but sounds too blunt. Or too cautious. Or too informal for the audience. AI can take the same content and reframe it (more diplomatic, more direct, warmer) while keeping the substance intact. This is hard to do well by hand, especially when you're close to the content, and AI handles it well.
Tightening and shortening. Most professional writing is too long. Not because the writer is waffling, but because cutting is harder than writing and there's rarely time to do it properly. AI is good at taking a 400-word email and turning it into a 200-word email that says the same thing. It's also good at identifying the sentences that aren't pulling their weight: the ones that repeat a point already made or add qualifications nobody needs.
Generating first drafts from context. When you're staring at a blank page and know roughly what you want to say but can't quite get started, a chatbot can give you something to react to. The first draft is almost never the final version, but it gets you past the hardest part: the beginning. This is particularly useful for recurring formats like job descriptions, meeting agendas, stakeholder updates, and project summaries. Once you've prompted well a few times, you develop templates and patterns that speed up every subsequent attempt.
Catching what you've stopped seeing. Everyone has blind spots in their own writing. Phrases you overuse, grammatical habits you've never noticed, sentences that made perfect sense to you but don't to anyone else. A writing assistant catches these mechanically and consistently, in a way that re-reading your own work a fifth time rarely does.
Adapting for audience. The same information often needs to be communicated differently to different people. A briefing for a board of directors reads very differently from an update for frontline staff, which reads very differently from a social media post. AI can take a single piece of source content and reshape it for multiple audiences, saving you from writing the same thing several different ways from scratch.
Where it falls short
AI writing tools have real limitations, and being honest about them makes you a better user.
Voice and personality. AI-generated text has a particular flavour. You may have noticed it already. It tends toward the smooth, the comprehensive, the slightly generic. It uses certain phrases more than a human would. Good prompting helps, but if your writing has a distinctive voice, or if personality is part of the point (a blog post, a fundraising appeal, a social media presence), AI-generated text often needs significant editing to sound like you rather than like a capable but anonymous professional.
Nuance in sensitive contexts. A redundancy notification. A complaint response. A message to a bereaved colleague. These are situations where every word carries weight and the wrong tone can do real damage. AI can draft these, and sometimes its suggestions are useful as a starting point, but the judgment calls are yours: what to say, what not to say, how to say it. Over-relying on AI for sensitive communications is one of the easiest mistakes to make, because the output sounds polished and professional even when it's missing something important that only you would know.
Knowledge of your specific context. AI doesn't know your organisation's culture, your relationship with the recipient, the history behind a project, or the politics of a situation. It can only work with what you tell it. For a routine email, this doesn't matter much. For anything where context is critical, you either need to provide that context in your prompt (as you practised in Thing 3) or accept that the AI's draft will need substantial reshaping.
Over-editing. There's a subtler risk with AI writing tools: the temptation to polish everything until it's flawless, which can strip writing of the qualities that make it effective. A perfectly grammatical, perfectly structured email is not always the best email. Sometimes directness matters more than elegance. Sometimes a slightly rough edge is more authentic than a buffed surface. The writing assistant will always suggest improvements (that's its job) and learning when to ignore those suggestions is as important as learning when to accept them.
A word about transparency and detection
One question that comes up often in professional settings: should you tell people when you've used AI to help with your writing?
There's no single answer, and practices vary across organisations and sectors. But a few principles are worth keeping in mind.
Using AI to check your grammar and improve your phrasing isn't meaningfully different from using a spell-checker, asking a colleague to proofread something, or consulting a style guide. These are tools that help you communicate better, and nobody expects you to disclose them.
Using AI to generate a first draft that you then substantially edit and make your own is closer to how many people use templates or adapt previous documents. The final product reflects your judgment and knowledge, even if the raw material came from somewhere else.
Using AI to produce a finished piece of work that goes out essentially as the AI wrote it, without meaningful review or adaptation, is different. This is where transparency matters most, because the reader is attributing the knowledge, judgment, and perspective in that writing to you.
The practical guideline emerging across most professions: use AI freely as a tool in your process, but make sure the final output reflects your knowledge and judgment. Be transparent about AI involvement when the context calls for it. If your organisation has a policy on this, follow it. If it doesn't, this might be a conversation worth starting.
Will people know?
Quite possibly, yes. AI detection tools have improved a lot, and the best of them, such as Pangram, are now good at recognising text that has come straight out of an AI tool. Organisations use them in recruitment, education, publishing and grant funding. And people notice too: writing that's polished but generic, with no personal detail or real opinion, tends to stand out.
Consider a job application. If you ask AI to write your cover letter and send it with only light changes, there's a good chance it will be recognised for what it is, either by a detection tool or by the person reading it. That matters because the reader wants to know about you: your experience, your judgement and how you would approach the role. The same goes for funding bids, tender responses and anything else where the reader cares whether the words and ideas are genuinely yours.
Detection is also moving beyond guesswork. Some AI providers now add invisible watermarks to the text their tools produce. Google, for example, uses a technology called SynthID to mark text from Gemini, and EU rules that have applied since August 2026 require providers to make AI-generated content identifiable by machine. On top of that, detection tools keep getting better, and anything you submit today can be checked again with tomorrow's tools. So whatever you might think now, it's safest to assume that AI writing passed off as your own will be spotted at some point.
So the aim isn't to avoid being detected. It's to make sure the work is authentic. Use AI as part of your process, for generating ideas, creating a starting structure or suggesting phrasing, and then write the final version yourself, in your own voice, drawing on what you actually know and think. Be open about how you used AI when the context calls for it. Writing that's genuinely yours is better writing, and it's honest.
As a rule of thumb: if you couldn't explain or defend every sentence in the text, you haven't made it yours yet.
Resources to explore
Free tier available with browser extension, desktop app, and mobile keyboard. Covers grammar, spelling and a monthly allowance of AI prompts. The browser extension works across most web-based writing tools.
A practical walkthrough of Grammarly's current AI capabilities, including the generative features and tone adjustment tools.
Overview of Gemini's integration with Gmail, Docs, and other Google apps. Useful if your organisation uses Google Workspace.
Apple's guide to using the system-wide writing tools on iPhone. Similar features are available on iPad and Mac.
Activity: the rewrite exercise
You're going to take a piece of professional writing and run it through two different kinds of AI writing assistance. The first is a writing assistant that works inside your editor: Grammarly, or Copilot in Microsoft Word or Outlook if your organisation provides it. The second is a chatbot rewrite using the conversational approach. You'll compare the results and develop a practical sense of which approach works best for which kind of task.
- Write something to work with. You need a piece of professional writing, at least a few paragraphs (ideally 150 to 400 words). Write something fresh for this exercise rather than using an actual work document. Your employer may not be comfortable with real work content being pasted into AI tools, and creating something yourself gives you a better baseline for judging what the AI changes. Write a realistic piece of professional communication based on an invented but plausible scenario: an email explaining a service change, a briefing summarising a project outcome, a message introducing a new way of working, a job description for a role you know well, or a response to a complaint. Make it feel real. Don't worry about making it perfect; the rougher it is, the more you'll see what the tools change. Save a copy of your original.
-
Run it through a writing assistant. Choose one of these two options:
- Grammarly (free plan). Install Grammarly's free browser extension from grammarly.com if you haven't already. Paste your text into a Google Doc, an email draft or Grammarly's own editor, which you can open once you've signed in at grammarly.com. Review the suggestions. Accept the ones you agree with and dismiss the ones you don't. Grammarly's free plan includes a monthly allowance of AI prompts. Highlight a paragraph, open Grammarly's AI assistant and ask it to make the text more formal, more concise or more confident. (One-click tone adjustment is a paid feature, so don't worry if you see it locked.)
- Copilot in Microsoft Word or Outlook. If your organisation provides Microsoft 365 with Copilot, paste your text into a new Word document or a new Outlook email draft. Ask Copilot to review it, then highlight a paragraph and ask Copilot to make it more formal, more concise or more confident. Only use the made-up text you wrote for this exercise, never real work documents or emails, and follow your employer's AI policy.
Save the improved version.
-
Try the chatbot approach. Open your preferred chatbot and paste in your original text (not the writing assistant version). Give the chatbot a prompt that includes context about what you're trying to achieve. For example:
Then try at least one more variation. Ask for a version that's more formal, or more direct, or restructured with the most important information first. Use the iterative skills from Thing 2: push back if something doesn't work, ask for alternatives, tell the chatbot what you liked and what you didn't. Save your preferred chatbot version.Here's an email I've drafted to a partner organisation about a change in how we deliver our outreach service. The tone needs to be professional but warm. We want them to feel reassured, not alarmed. Can you rewrite this to be clearer and about 30% shorter? Keep the key facts but make it easier to scan quickly.
- Compare all three versions. Put your original, the writing assistant version and the chatbot version side by side. For each, consider: What improved? What was lost? Which feels most like you? Which was most useful for this particular piece of writing? How much editing did you need to do?
Your output: a document containing your original text, the writing assistant version (from Grammarly or Copilot), the chatbot-rewritten version (your preferred one), your annotations comparing all three, and a few sentences on when you'd use each approach in future.
Claim your Open Badge
Your submission should include your original text, the two AI-assisted versions, and your comparative annotations.
Submit your original text, both AI-assisted versions, and your comparative annotations as evidence to claim this badge via cred.scot.
Claim now