Here's something I'm GENUINELY hyped about.
Every software company on earth is trying to get you to come to them. Open our app. Log into our dashboard. Live in our tab. That's the whole game, has been for years, get the user inside your four walls and keep them there. And I get why. But it means every tool is quietly pulling in the same direction, all tugging at your attention, all wanting to be the place you spend your day.
We just did the opposite, and I'm properly excited about it.
Instead of building yet another reason for you to come to us, we packed up our intelligence and sent it to where you already are. That's what shipped this week, and it's my favourite thing we've built in a long time.\
Quick bit of context. Nobody opens a marketing dashboard for the joy of it. Your team doesn't live in analytics tools, they live in Claude and ChatGPT, writing copy, drafting briefs, building landing pages, all day. And that AI they're working with? Knows nothing about your business. Blank slate, every single time. You paste in the context, it forgets by Tuesday, you paste it again. Everyone's just quietly accepted that as normal.
So we killed it. It's called Connected AI, and the short version is your team's own AI tools can now read your Brain directly. Your goals, your voice, last month's numbers, your reports. "Write me a landing page" in your Claude now starts from your actual customers and your real performance, not a guess. One connection covers every tool, because it runs on MCP, the open standard that's become the way AI tools talk to data. A universal port. Plug the Brain in once, every assistant can see it.
But here's the part I actually want to talk about, because reading your data is the easy bit and half the market does it already.
Ours can learn.
There's a setting we called Teach, and it runs the other way down the pipe. Someone on your team tells their AI something true, a shift in your positioning, a nugget from a client call, the answer to something the Brain had been wondering about, and the AI can save it straight back into the Brain. Your intelligence gets sharper from work that was happening anyway. No dashboard. Nobody sat there feeding the machine on a Friday afternoon. They just did their job and the Brain was paying attention.
I'll be straight about why that one took us the longest. Letting an AI write into your single source of truth is a scary thing to build. The nightmare's obvious, the robot confidently saves something wrong and rots your data from the inside. So we didn't let it. Confident stuff gets saved with a full trail you can undo. Anything it's unsure about doesn't get written at all, it waits for a human to say yes. Nothing gets rewritten behind your back, and every move is logged where you and we can see it. We built the paranoid version deliberately, because a Brain you can't trust is worse than no Brain at all.
Small thing I'm weirdly proud of, too. When your AI pulls a number, the caveats come with it, how fresh it is, whether it's an estimate, how solid the attribution is. So your assistant won't hand you a wobbly figure like it's gospel, or set off your alarm bells about a dip the Brain already understands.
Setup's about a minute. Sign in with your Claude or ChatGPT login, tick the permissions on a screen that shows you exactly what you're handing over, done. Cursor and the developer tools use a key instead, same thing, different door. Every connection's named and revocable, so if a laptop walks or someone leaves, you cut that one and everything else carries on.
It's rolling out to self-serve workspaces first. If it's landed in yours, it's in Settings under Connected AI right now.
Ten years of everyone building walls to keep you in. We just opened a door and walked our best thing over to your side of it.