Welcome to this edition of Loop!

To kick off your week, I’ve rounded-up the most important technology and AI updates that you should know about.

‏‏‎ ‎ HIGHLIGHTS ‏‏‎ ‎

Let's jump in!

This guide distills 10 AI strategies from industry leaders that are transforming marketing.

Outpace your competitors by mastering AI.

Get Your Free Guide

1. Microsoft is bringing Anthropic’s Cowork into their platform

We start this week with Microsoft’s announcement of Copilot Cowork, which is an important step forward for automating daily tasks and simplifying the way that we work.

While the name doesn’t roll off the tongue easily, this is a close partnership between Microsoft and Anthropic - which integrates Claude Cowork with their enterprise tools.

Essentially, Cowork is a single platform that allows you to easily run tasks with AI agents. You can upload your own data files, provide them as context to the agent, and grant it access to specific tools - such as Outlook emails, Teams chats, or other Microsoft services.

This allows you to run tasks in seconds, with an AI agent that can create a plan for solving it, call different tools, and ask for your feedback. It’s a more advanced way to do work, when compared to simple chat interfaces - as you no longer need to paste every file yourself.

With Cowork, the AI agent can decide what info is needed and then search for it across your Microsoft 365 apps. For example, Cowork can reschedule your meetings, assemble a briefing deck from your emails and files, create company research from SEC filings and analyst reports, or coordinate a product launch across multiple documents.

Everything stays within Microsoft 365's security and governance limits, which is important for enterprise customers and will make it much easier for them to adopt the tool. It’s a clear sign that Microsoft is diversifying beyond OpenAI, which is something we've been seeing more of across the industry.

It’ll be interesting to see how this works in practice. I suspect there will be a big difference in how this is used internally by companies, versus how they use it with clients. Automating your team’s research work is one thing, but trusting an AI agent to reschedule meetings or send an email to clients is a completely different ballpark.

If your company is already deep into the Microsoft 365 ecosystem, this is worth following - but I'd wait to see how it performs outside of the demo environment before getting too excited.

2. UK Government invests £500 million in new sovereign AI fund

The UK government has launched a Sovereign AI Unit, which will be backed with up to £500 million ($661 million) in funding.

This new unit is tasked with reducing Britain's dependence on foreign hyperscalers - like AWS and Azure - and will invest in domestic compute, hardware, and data infrastructure.

The move is in response to growing unease across Europe, as US technology companies are deeply embedded in their economies and the current administration is threatening to use that leverage against its allies.

Just to put the sheer scale of this into perspective, US technology companies - including Amazon, Microsoft, and Google - control more than 70% of the European cloud market. There are some European competitors, but they lack the scale and depth of their American competitors.

While the UK’s £500 million figure sounds large, it's pretty modest compared to the tens of billions that US and Chinese governments are pouring into sovereign compute - which raises the question of whether this is enough to build genuine independence or just enough to make a political statement.

The real test will be execution and getting public sector organisations to migrate away from the US tech giants, which will be difficult. There’s no point funding domestic companies and then ignoring them when it comes to government procurement.

Regardless, it’s a good move from the UK. There are rising tensions with the US over Ukraine, Greenland, and fresh economic tariffs - so the country desperately needs to invest in its own domestic tech sector. It’s been a long time coming, but it’s better late than never.

3. Google Maps is rebuilt with AI support and is much more interactive

The company has announced a major upgrade for Google Maps, which now has their Gemini AI model built-in and a more immersive experience for drivers. Previously, finding local information meant sifting through reviews and piecing together answers from multiple sources.

Now you can just tap "Ask Maps" and type questions like "where can I charge my phone without waiting in a long queue?" - and Gemini will pull info from 300 million places and reviews from 500 million contributors to give you an accurate answer.

The feature goes well beyond simple queries, too. You can ask something like "I'm headed to the Grand Canyon and Horseshoe Bend - any recommended stops along the way?" and get directions, ETAs, and insider tips from other tourists.

The Ask Maps feature is a really interesting upgrade, as it shows how Google is embedding Gemini into products that people actually use every day - rather than simply trying to sell another standalone AI tool.

The personalisation angle is really clever too, as Gemini can access data about your dietary preferences, your saved places, and your search history, and then adjust its suggestions.

Of course, that does raise the obvious question of just how much behavioural data Google is feeding into these responses, and whether users will trust a chatbot for decisions like restaurant bookings remains to be seen.

Ask Maps is live now in the US and India on mobile, with Immersive Navigation rolling out across the US over the coming months.

4. How LangChain built an advanced AI Agent for their sales team

I thought this was an interesting article from LangChain’s development team, as they shared some of the key learning from building a GTM agent.

The company has built a new AI agent, which is able to automate the entire sales outreach process - from lead research to email drafting - and the results are pretty impressive. They’ve seen a 250% increase in conversion and their sales reps have been able to reclaim 40 hours of work per month.

To do this, the agent is triggered when a new Salesforce lead comes in, it then checks whether outreach is appropriate - including whether a teammate has already been in contact - and then pulls context from Gong, LinkedIn, and the web about that sales opportunity.

Once the agent has everything it needs, it then drafts a personalised email and waits on the sales team to approve it. Nothing is sent without human approval, which is the right call since a poor, automated email can easily damage that relationship with prospects.

But what makes this more than your typical automation story is the learning loop. When LangChain’s sales reps edit the draft emails, their system is able to analyse the difference and then logs what the preferred writing style is for each rep. This allows it to improve the future drafts over time, without the sales rep having to take any action.

It’s an interesting blog that covers how LangChain built this agent and the design choices they made, which isn’t something that most companies publicise. If you’re currently building your own AI agents, this is worth sharing with your team.

5. Anthropic launches a tool to combat the rise of AI-generated code

The startup has launched Code Review, a new feature in Claude Code that uses AI to automatically review pull requests before the code is merged into your codebase.

This new product is aimed squarely at enterprise customers - companies like Uber, Salesforce, and Accenture - who are already using Claude Code and are finding that the sheer volume of AI-generated code has created a review bottleneck.

The system integrates with GitHub, can automatically analyse the pull requests, and adds comments that flag potential issues.

Interestingly, Anthropic has deliberately focused on logic errors rather than styling issues. It’s a solid move, as most developers have learned to ignore AI feedback unless there’s something they can immediately change.

Under the hood, it uses a multi-agent architecture where several agents will examine the codebase from different angles. A final agent then collects these issues and ranks them by severity.

Personally, I think the positioning is smart, but the pricing is quite high. At $15 to $25 per review on average, this only makes sense at a huge scale. For large enterprises that are shipping hundreds of pull requests a day, the maths probably works out, but for smaller teams it could add up quickly.

The timing is also worth noting - Anthropic is leaning heavily into enterprise revenue as it navigates its ongoing legal dispute with the Department of Defense, with Claude Code's run-rate revenue is now reportedly surpassing $2.5 billion.

ChatGPT and Claude can now help you learn new topics

Both OpenAI and Anthropic have upgraded their chatbots in recent days, allowing them to generate charts, diagrams, and interactive images directly within your conversation - rather than in a separate side panel.

This seems to be part of a wider push into education, which is something that people have underpriced in recent years. For a long time, I’ve been bullish about how the technology can be used to educate young people and allow them to better understand topics.

Whether you’re trying to learn how interest compounds over time, or understand science topics like Charles’ law, these AI models can create interactive tools that help you visualise what’s going on.

Of course, AI models do have their problems and there is the risk that it makes a mistake. There’s not much point in explaining mathematical equations with diagrams, if those diagrams are completely wrong.

But I do feel these hallucinations are becoming less frequent, so the risk is reducing over time as the models become more advanced. It’s just worth keeping in mind, as the technology does still have limitations and won’t always be perfect.

If you want to try this out, you should already have access to the new feature. Just ask your chatbot to explain a topic and generate interactive diagrams that clearly illustrate it.

How to create a short film with AI

Last week, I presented at BelTech - Ireland’s biggest tech conference - and talked about how AI tools are changing the way we create movies.

Despite having no filmmaking experience, I was able to create a short animated film and then presented it back to the audience at BelTech.

There was a great response to the talk, so I’ve turned it into a free guide that walks you through the workflow - allowing you to create your own short film.

It includes the entire journey, from brainstorming ideas with Claude, to generating video clips with Google Flow, to adding sound effects with ElevenLabs.

I shared a link to the guide in last Thursday’s email, but I’m resharing it again in case you missed it. You can read the guide using the link below.

🪖 US Army announces new contract with Anduril worth $20 billion

🧑‍💼 Meta considers layoffs that could affect 20% of the company

🎨 Canva’s new editing tool adds layers to AI-generated designs

📱 A quarter of iPhones are now made in India

🧠 Yann LeCun's startup raises $1.03 billion to build world models

🚁 Electric air taxis are about to take flight in 26 states

🚕 Uber, Wayve, and Nissan plan to launch robotaxis in Tokyo this year

🌧️ Google is using old news reports and AI to predict flash floods

📺 YouTube now generates more ad revenue than Disney, NBC, Paramount, and WBD combined

🕸️ Meta acquires Moltbook, the viral social network for AI agents

This startup is building autonomous agents for the cybersecurity industry and has just raised over $190 million from investors. The company was founded by Kevin Mandia, who previously led another cybersecurity company called Mandiant and was sold to Google for $5.4 billion.

Notably, the CIA's venture arm - called In-Q-Tel - has invested in the company and they seem in a great position here. With the rise of AI-powered cybersecurity attacks, defenders need their own autonomous agents to keep up.

Mandia believes that autonomous AI hackers will be able to complete attacks in minutes that currently take days, which is a bold claim but not an unreasonable one given how quickly the offensive AI tools are advancing.

His co-founders are all former Google and Mandiant engineers, so the team has a strong background in the cybersecurity sector.

Of course, building autonomous defensive agents is far harder than it sounds - the margin for error in security is essentially zero, and an agent that misidentifies threats or misses edge cases could be worse than no automation at all.

But with Mandia's track record, serious backing from defence-adjacent investors, and a founding team that has actually built security infrastructure at scale, Armadin is certainly one to watch.

This Week’s Art

We’ve covered quite a bit this week, including:

If you found something interesting in this week’s edition, please feel free to share this newsletter with your colleagues.

Or if you’re interested in chatting with me about the above, simply reply to this email and I’ll get back to you.

Have a good week!

Liam

Share with Others

If you found something interesting in this week’s edition, feel free to share this newsletter with your colleagues.

Liam McCormick is a Senior AI Engineer and works within Kainos' Innovation team. He identifies business value in emerging technologies, implements them, and then shares these insights with others.