Also - totally agree with the quote about “living in a parallel universe”. I've spent many, many hours over the last few weeks buried in Claude CoWork - it is a totally different experience and approach to interacting with a normal AI chat bot. As with getting the best out of Claude Code, the more thought and planning you put in up front, the better the results. Figuring out the best to way brief CoWork and which skills, agents, plug ins etc should be assembled to go and do the work (or just getting Claude to assist you with making those decisions). Also, just pointing it at the right folders and saying the necessary background you need is all here (you can also now add existing Projects as additional context).
Someone asking MS CoPilot to summarise an email for them has no idea of what someone with expertise and experience armed with access to something like Claude CoWork can do - and it is only been out for few weeks - wait till the Window version launches - can't be long now. This really does feel like a significant "moment".....
A moment, indeed! Windows versions of Cowork won't help the poor souls denied access to it, but if it carries on its current course, we'll see a lot of MSFT shops take it on as approved.
Building one's toolkit of data and knowledge through Claude Skills, Plugins and the like is going to be liberating for those who can get their hands on them. And they port easily to other systems so - for now - you can avoid getting locked in to any stack. In an act of recursive nerdery I used Claude Code to develop my own CLI tool with a different kind of memory system. I'm plugging in Opus 4.6 to power it but, there's no reason I couldn't swap it out for Mistral or Chat or whatever.
Fabulous observations, Anthony. You articulated something that I've been unable to put straight in my own mind.
This idea of restricted opportunity through IT compliance, is tantamount to compliance for self-preservation rather than innovation. It reminds me of the stories of market leaders who are scared to disrupt their business model or even their product range with anything dramatically innovative or different because the cash cow keeps producing the cash.
All of a sudden at some point another competitor comes along with something radical or different in their space and everybody flocks towards that.
Apply that to AI and you have these companies that are restricted by their compliance to an enterprise version of AI who are unquestionably going to be at risk of being attacked or even isolated by an AI native startup that can do exactly the same thing or even better with a fraction of the resources, a fraction of the people.
I look at the project I was involved in at my last role at Microsoft. We were trying to find a simple and easy way of summarising and capturing a plethora of regulatory changes such that our lawyers could quickly make the appropriate judgments, and our engineers implement the appropriate specification changes to ensure that the product was compliant.
A simple plug-in or an update from Claude has managed to replace an entire team. While the team was long gone for many different reasons before the Claude plugin came out it’s probably the best example of where an AI native startup can look at the same problems as the big businesses and run roughshod over the larger organization.
I would not be surprised if within the next year to eighteen months AI native startups are producing tools that are not only innovative, cost-effective, and compliant but meaningfully impact major industries where compliance is almost considered protectionism.
Thank you for adding your thoughtful and insightful comment, Paul. I feel like it should be a post of its own.
I know compliance, regulation and business systems will all change and I wonder how quickly in each sector. Software and services feel like they will move faster – but how fast is going to be the thing to watch.
I can’t help feeling we’ll see a number of AI-native startups appear that address specific industry issues or processes, and instead of needing to build enterprise-integration services, agents will do the work for them.
It’s fascinating, rather like watching investment in optimising the size and shade of red the flags that had to be carried in front of early motor vehicles. This systemic, and Donella Meadows was right- it’s about changing paradigms, not tinkering with obsolescence.
Also - totally agree with the quote about “living in a parallel universe”. I've spent many, many hours over the last few weeks buried in Claude CoWork - it is a totally different experience and approach to interacting with a normal AI chat bot. As with getting the best out of Claude Code, the more thought and planning you put in up front, the better the results. Figuring out the best to way brief CoWork and which skills, agents, plug ins etc should be assembled to go and do the work (or just getting Claude to assist you with making those decisions). Also, just pointing it at the right folders and saying the necessary background you need is all here (you can also now add existing Projects as additional context).
Someone asking MS CoPilot to summarise an email for them has no idea of what someone with expertise and experience armed with access to something like Claude CoWork can do - and it is only been out for few weeks - wait till the Window version launches - can't be long now. This really does feel like a significant "moment".....
A moment, indeed! Windows versions of Cowork won't help the poor souls denied access to it, but if it carries on its current course, we'll see a lot of MSFT shops take it on as approved.
Building one's toolkit of data and knowledge through Claude Skills, Plugins and the like is going to be liberating for those who can get their hands on them. And they port easily to other systems so - for now - you can avoid getting locked in to any stack. In an act of recursive nerdery I used Claude Code to develop my own CLI tool with a different kind of memory system. I'm plugging in Opus 4.6 to power it but, there's no reason I couldn't swap it out for Mistral or Chat or whatever.
Agree on the portability aspect - who knew well-crafted markdown files could provide so much value 😀
Fabulous observations, Anthony. You articulated something that I've been unable to put straight in my own mind.
This idea of restricted opportunity through IT compliance, is tantamount to compliance for self-preservation rather than innovation. It reminds me of the stories of market leaders who are scared to disrupt their business model or even their product range with anything dramatically innovative or different because the cash cow keeps producing the cash.
All of a sudden at some point another competitor comes along with something radical or different in their space and everybody flocks towards that.
Apply that to AI and you have these companies that are restricted by their compliance to an enterprise version of AI who are unquestionably going to be at risk of being attacked or even isolated by an AI native startup that can do exactly the same thing or even better with a fraction of the resources, a fraction of the people.
I look at the project I was involved in at my last role at Microsoft. We were trying to find a simple and easy way of summarising and capturing a plethora of regulatory changes such that our lawyers could quickly make the appropriate judgments, and our engineers implement the appropriate specification changes to ensure that the product was compliant.
A simple plug-in or an update from Claude has managed to replace an entire team. While the team was long gone for many different reasons before the Claude plugin came out it’s probably the best example of where an AI native startup can look at the same problems as the big businesses and run roughshod over the larger organization.
I would not be surprised if within the next year to eighteen months AI native startups are producing tools that are not only innovative, cost-effective, and compliant but meaningfully impact major industries where compliance is almost considered protectionism.
Thank you for adding your thoughtful and insightful comment, Paul. I feel like it should be a post of its own.
I know compliance, regulation and business systems will all change and I wonder how quickly in each sector. Software and services feel like they will move faster – but how fast is going to be the thing to watch.
I can’t help feeling we’ll see a number of AI-native startups appear that address specific industry issues or processes, and instead of needing to build enterprise-integration services, agents will do the work for them.
Great post - and I see similar with orgs I deal with - I wonder what the “ clunky-but-compliant AI chatbot" you refer to is?😀
Stay tuned for more revelations.
It’s fascinating, rather like watching investment in optimising the size and shade of red the flags that had to be carried in front of early motor vehicles. This systemic, and Donella Meadows was right- it’s about changing paradigms, not tinkering with obsolescence.
She was! Thank you for the connection to that thought.