We have been quiet for a while. So, what did we actually build? We went deep into the infrastructure to create Zaro’s new Guardrails Engine. This ensures that every piece of data entering our system is completely AI-ready for right now, and well into the future. Here is what that looks like in practice for your workspaces: • Smarter Agents, Safely: Agents can now read and write back to the same context layer, building on their previous interactions, while everything remains locked down and secure. • Strict Clearance: Agents only see the information they are explicitly cleared for. No exceptions. • Total Transparency: Every single agent leaves a trace that can be tracked. • True Data Ownership: Every piece of data stays yours. Period. Take it with you whenever you want. Every team. Every agent. Governed end-to-end. This is the context that you finally control. Check out the video to see it in action, and dive into the nitty-gritty details over on our website - link in the comments.

We started Zaro.ai because we believe AI can do much more for a business with one current, governed account of how it works, one that people and AI can both use and contribute to. When a decision or correction becomes part of that shared context, the next project can build on it rather than beginning again. The chance to put that belief to work with @juventusfc is extraordinary. This is a moment to appreciate. Then back to building.

Situational update: not again please 🤦🏼‍♀️

We ran a hackathon with @awscloud (AWS) last Friday. The brief was simple: bring a real problem from your own job and leave with the agent(s) that solve it. Everyone did. Nobody needed help with building. Twenty minutes in, we showed what becomes reachable once your data is AI-ready and governed. Examples of tools we use in Zaro: what the agent can do when it can read everything it needs to, write back to it, and be scoped to only the parts it's allowed to touch. You could watch people's first agent become more specific as soon as they'd seen it. The brief they'd walked in with was almost always smaller than what was possible. Most people built tools around document processing. Invoices, onboarding paperwork, compliance files, and sharing context but not sensitive information between working departments. None of these is considered 'exciting' use cases, which tells you how big the gap is between what's possible with AI and what's delivered. That's the real constraint. Not whether a model can do the task, but whether your data sits somewhere an agent can read, write back to, and the next one can pick up where the last one left off. Tomorrow we're publishing the longer answer to exactly that: what "agent-ready" means for enterprise data and workflows, as a practical checklist. Thanks to ShallWe.Tech and Xiaowei Zhu for hosting.

Last Thursday we hosted our first AI Transformation Table at the Quilon restaurant in London. 17 leaders in the room, one seat per company. Everyone there is actually spearheading AI inside a large organisation, which makes for a very different conversation than the ones happening online. The thing that came up again and again: pointing AI at your existing infrastructure doesn't get you the value that was promised. It starts with data and context. And that layer can't be a one-off migration; it has to keep moving at the speed of the business, or it becomes another failed project. Most people at the table had learned that the hard way. Thanks to everyone who came. The next one is in October. DM us if you'd like a seat.

Your workspace no longer stops at your desk. Zaro is now live on the App Store, available to all existing users. Every workspace, file, app, and agent. Same guardrails, same access controls, now in your pocket. Click the link below to download for free 👇👇: apps.apple.com/gb/app/zaro-a…

Kings Cross/St Pancras has been evacuated today. Here's the view from our office. Back to work now; nothing will stop us building Zaro

Anyone heard that big bang near King's Cross/St Pancras just now? 😳

King's Cross is fast becoming one of the best places in the world to build an AI company. @GoogleDeepMind , @OpenAI, @xAI, and @Meta are all here. @AnthropicAI is not far away. Plus all the newly emerging frontier labs. We're happy to be building Zaro here. Right at the heart of it. Our biggest complaint is the mobile service. There literally isn't any. It makes no sense. This is what our CTO has to do whenever he gets a call. Can @KanishkaNarayan, or @UKSovereignAI , do something about this please...?

Everyone keeps asking us which model to use. Wrong question. We've spent the last few months talking to enterprise teams about their AI plans. Almost none of them are blocked on the model. What they actually want is simple to say and very hard to build: their Slack, their email, their Snowflake, their BigQuery - in one place. Somewhere they can query it, visualise it with real apps, and let agents act on top of it. Instead, their data sits in ten systems that were never designed for an agent to touch. Same three complaints, over and over: MCP and APIs - too slow, too much glue code, breaks the moment you need real-time. Knowledge bases - fine for retrieval, useless the moment an agent needs to update something. Knowledge graphs - nobody wants to maintain one. So teams build a chatbot, demo it, everyone claps, and then it dies. A system that can only read can't do anything. This is why we built Zaro the way we did. One context layer across your siloed tools that agents can write to as well as read - versioned, governed, audited. Your agent updates context without corrupting it, and you can see who changed what. AI transformation isn't buying 5,000 seats and hoping MCP holds. It's getting your data into a shape AI can actually work with. If your agents can read your data but can't safely update it - how are you handling it?