Hacker News new | past | comments | ask | show | jobs | submit
They don’t get so much attention on HN, where the focus is naturally on coding, but the new Claude features being steadily rolled out for office work could, it seems to me, become a real moat. Anthropic has lately announced integrations of Claude into whiteboards, online meetings, etc. As people get used to using those and to having Claude as a collaborator who is knowledgeable about their work, they will become reluctant to switch to other providers.
But we already get this with copilot today, its integrated into sharepoint, teams, outlook, word, powerpoint etc. Is it a bit rough around the edges... yes, but you can finally find documents in sharepoint, where before that basically did not exist. We already have contracts with microsoft in regards to data protection, and Anthropic has not really done anything to prove that they can be trusted with it yet.
We'll also get this with tool use and browser automation. The models are increasingly good enough that they can build custom integration tailored to your specific use case. It'll be very hard to maintain a lead there when your "lead" is functionality the models can write unilaterally on the whim of a user.
loading story #50558148
Yes and that's great for the old enterprise organisations that are in the Microsoft ecosystem.

Is a 2 person startup going to choose Microsoft Teams and Sharepoint? I very much doubt it.

They can all coexist but they are for different target customers.

Your theory hinges on the idea that Claude users will subscribe to the idea of it being a knowledgeable collaborator and have that work for them.

I'm not saying that's impossible, but it's entirely speculative at this point. Developers are far more effective with these tools because they understand their strengths, limitations, and how to use them to get good results.

I've yet to see anyone benefit from treating the tool as a personified collaborator. They either scale back their use or fall flat on their face when it behaves unexpectedly.

I'm using M365 through MCP/CLI via some GraphQL interface, with access to Teams, documents, Outlook, etc. None of that depends on Anthropic.

I have non-dev clients and business partners that are all coding & mailing and doing everything with AI. Some I helped, but some are just handy. Next year goign to be insane.
This is exactly what they should be doing, yet I'm still baffled by their limitations. I can only connect one email account or one calendar to my Claude. What if I have a personal life and a job? Sorry, can't do it.

ChatGPT doesn't have this issue. It doesn't seem like it's particularly complicated.

That's assuming you're letting these integrations near your data. I recently had a look at the Claude plugin for Chrome and it honestly scares me a bit. This thing is a security nightmare (plus, probably a walking GDPR violation in the EU). They have solved neither the lethal trifecta nor prompt injections nor even unintentional misbehavior; yes, they have some mitigations and safeguards in place, but these cannot be remotely considered guarantees.