Using Claude Cowork for Grant Writing

Claude Cowork is a great tool, allowing you to work on a task in a permitted folder of documents, where the app can add or edit documents.

It's genuinely impressive. But it's early days. And there is a limiting factor when it comes to using it with clients - how it handles your data.

In this post:

• How is Cowork different to Chat?
• Getting set up with Claude Cowork
• Why Cowork is good for grant writing
• The ways Claude Cowork is not good for grant writing

How is Cowork different to Chat?

The main difference to an LLM is that in Cowork you work towards a goal rather than asking questions. You can connect it to a folder and external tools like calendars and email. So it can pull in the information it needs rather than asking you to supply everything manually.

You can delegate tasks to it and watch its thinking as it progresses. And it often stops to ask you questions when it needs more direction, sometimes giving you multiple choice options.

You can build a great context architecture around your resources, using grounded data.

Getting set up with Claude Cowork

You need a paid account for Claude Cowork, but it's a small amount for the value you get from it. You also need the desktop version of the app for Cowork.

When signing up, it's worth bearing in mind that you can't change the email you sign up with. You have to delete your account and create a new one if you want to do that.

In settings - turn off model training. Give it limited access to start with to documents without any sensitive information. Don't give it access to your whole documents folder. Create a dedicated folder you pull docs into.

Then, the simplest thing to do is set up an account and ask Claude to walk you through it.

One thing I found was that I couldn't find the button for allowing folder sharing - but you can say to Claude Cowork you want to do it and it shows you the following:

You can also combine Claude Cowork with Obsidian for a great grant writing workflow. I've written a post about getting started with Obsidian.

You have a choice of models, with the most advanced thinking model taking longer over the task. But the more powerful the model you choose, the faster it eats your usage limit - and once you do that it locks you out of the app (we're talking hours, which isn't great if you have a deadline.)

One useful thing to do is assign a key on the keyboard for dictation, like control pressed twice - makes it much quicker than typing.

Why Cowork is good for grant writing

The main reason Cowork is useful, I think, is that it allows you to build a library of resources for your grant writing, which you can access and analyse in the app.

Policies, context, reading around the subject… you can interrogate it all easily in Cowork.

For example, one of the docs might be the UK Government's Guide to developing the Project Business Case. As the data is grounded, you can ask whatever questions you want about, say the Five Case Model. You can find themes between different policies from different organisations. In short, it allows you to handle a large amount of data with ease and helps your thinking.

I wouldn't use it to write anything, though. You get that deathless LLM phrasing. And it's hardly a way to stand out from the crowd. What added value are you bringing if you do that?

The ways Claude Cowork is not good for grant writing

How Claude Cowork handles your data at the moment is sub-optimal, especially from a company that referred to itself as safety first when it first introduced Claude in 2023. In February 2026, according to TIME Magazine, Anthropic dropped this claim.

In 2023, Anthropic committed to never train an AI system unless it could guarantee in advance that the company’s safety measures were adequate. For years, its leaders touted that promise—the central pillar of their Responsible Scaling Policy (RSP)—as evidence that they are a responsible company that would withstand market incentives to rush to develop a potentially dangerous technology.

How does that manifest in the product? You can't easily delete your data, is how.

You can delete conversations in Chat, and then they are fully deleted from the Anthropic servers in thirty days.

However, with Cowork, you can only archive your Cowork sessions or projects. To delete them you have to email Anthropic to ask for it to be done, which is unworkable.

And so, you are better off working on more sensitive documents in chat, which makes cowork… a little less useful than it could have been if they had shipped it with better privacy controls.

All that being said, it's still mightily impressive.

Tiago Forte has a useful YouTube video showing it in action - What Claude Cowork actually does, and why it's different.