On the night of 14 April 1912, the RMS Titanic sank after hitting an iceberg. What's often left out of the story is that the ship received warning after warning.

Vessels across the North Atlantic reported ice. Wireless operators logged the messages. Some officers heard about the danger; other warnings were delayed or lost in routine traffic.

The iceberg wasn't really unknown.

The problem wasn't missing information. It was missing context. Nobody on board held the whole picture. The warnings arrived as separate fragments, and each was easy to wave off on its own. Put together, though, they were a clear and urgent signal.

I’m betting that your business has a bunch of big old icebergs in the water.

All the information you hold right now behaves like those warnings: recorded, filed, scattered. It's all "saved.", but never really there when you actually need it.

So teams repeat mistakes they already made, walk past risks they already spotted, and rebuild things they already built. Storing information is super easy; take your pick of tools. But getting the right thing to the right person at the right moment isn't close.

That context, the stuff in people's heads and in Slack thread, is your actual asset. It's why your senior people cost what they cost. It also walks out the door every time someone quits, or just takes two weeks in Crete.

So why has nobody fixed this? And what's changed?

Nobody could solve this before. You can't read every doc, every ticket, and every thread, keep it in your head, then hand the right piece to the right person at the exact second they need it. People don't work that way. Knowledge management has been thirty years of expensive software nobody opens.

AI breaks that limitation. A model can read across all of it and remember more than any human.

So every smart company is doing the same thing. Buy a tool, point it at Slack and the docs and the tickets, ask it a question, get back a clean, confident answer. Demo looks incredible.

Everyone nods.

Then it ships garbage.

What they built reads everything and understands nothing. It's a search index with a confident voice, but with no context.

Your data says what happened. The commits, the tickets, the calendar invites etc. But, none of it says what any of it meant.

The meaning lives somewhere else. There might have been the final ‘decision’, but the three options you killed to reach it are key info. And that's the info which separates an answer you can trust, with one that just doesn’t sound right. That info is sitting in notes, transcripts and Slack threads.

Loads of companies are building these right now to solve the context problem. They’re searchable memory for everything the business knows, connecting to all the different knowledge tools you have.

Brilliant, great start. And will be worth it’s weight in gold.

But a search box only works if you know what to search for. The junior engineer never looks up the migration that took payments offline two years ago. She has no idea it happened. The person who needs the lesson most is the one who doesn't know it exists.

Reactive AI saves you time, and is amazing when you know what you are looking for.

But what if we could make the tool proactive, so that it surfaces timely knowledge that you

But we believe that proactive tools can save you from yourself. But how do we make the AI proactive…?

So we built Roger: a proactive agent, with a shared memory across Planes.

Roger knows what the company knows: the decisions, the people, the risks, the plans, how the place actually runs.

Everything it tells you traces back to something written down (docs, notes, briefs), so it doesn't make things up. Work happens in meetings, on Slack, and in code, and Roger files it as it goes. Every project gets its own memory, plus one shared brain for the whole org.

Roger can be accessed in Slack, Linear, and GitHub. It remembers what you told it yesterday in a different tool. And it doesn't wait to be asked.

But Roger isn’t a chatbot stapled to a wiki. Roger uses its business rules and knowledge to guide the user without them asking.

Roger creates its own rules over time based on what it observes. How you scope, how you ship, what good looks like. Then it uses them to keep people on track while they work.

Then Roger starts tapping you on the shoulder.

You're a PM. You're about to roll out a billing change behind one feature flag. You hit send on the plan, and Roger Pings you.

Roger: "Heads up, you've got the new billing flow going out behind a single flag. Last time we did that, the flag stuck half-on for a week and we switched to a staged rollout with a kill switch. Want me to update the plan?"

You didn't ask, but this is news to you. Thanks Roger. Let’s say you’re scoping an app launch:

Roger: "Hey, I see on your plan that you have allowed three days for app store review. We take a week on average, so suggest you allow for that”.

Blimey. Cheers Rog!

We’re still in the early days of Roger, fine tuning things like the threshold for when Roger should comment (so he doesn't become annoying) and managing the impact of hallucinations. Like any ambitious product, we’re working through the nuances as we build.

But this highlights the difference between AI adoption and being AI native. Most companies are simply bolting AI onto what they already do; a copilot in the code editor or a summariser in the inbox. AI native means wiring your company so its own institutional experience shows up inside the decision, exactly while the decision is being made. A business that catches what it learns, connects it, and feeds it back gets sharper with every single project.

The Titanic's warnings were all written down. The ship just had no way to connect them in time. Now you do.

If you want to see Roger in action, join us on Friday 10 July for a webinar where we'll lift the lid on the AI native workflows we've been building, for ourselves and for businesses across financial services, law and healthcare. Sign up now.