I’m a Saturation Diver. A Legal Problem Led Me to Build BriefBase.

I work offshore as a saturation diver. Long rotations away, weeks at sea, and long stretches of time living in a hyperbaric chamber where you read, think, and end up exploring ideas you probably wouldn’t if life was busier. That’s how I first started going deep into AI.

But the reason I started using it wasn’t curiosity or hype. It came from a legal situation of my own.

Nothing dramatic, but complicated enough that I found myself going back and forth with a solicitor trying to explain the situation properly. There were emails, documents, explanations, and plenty of legal language involved. And something became obvious during that process.

What I thought my solicitor needed to know, and what they actually needed to know, were two very different things.

I would send long explanations trying to cover everything. Background, context, my interpretation of events. They would come back with more questions, trying to get to the key facts. We got there in the end. But looking back, if the situation had been structured clearly from the start it would have saved time on both sides. And probably money too.

That experience stuck with me. Because even after reading guidance, forums, and legislation, I still struggled with something surprisingly basic.

Explaining the situation clearly.

In my head it made sense, but when I tried to write it out or imagine explaining it to someone else, it always felt messy. Too much background. Details that might not matter. Important points buried in the middle.

So I started experimenting with AI.

At first it was curiosity. I’d type parts of the situation and see what happened. Sometimes the responses looked impressive, sometimes they were completely wrong.

And that was another important lesson. More than once the AI made confident assumptions that turned out to be incorrect. I caught them, but if I hadn’t known enough to question the answers those mistakes could easily have sent me in the wrong direction.

That made something very clear to me. AI on its own is not a reliable guide in legal situations. If the conversation isn’t structured properly it will happily follow the user down the wrong path.

That’s where the real light bulb moment came.

When guided properly AI wasn’t just producing answers. It could help organise the chaos. If I explained the situation in plain language and the conversation stayed focused on the right points, it could start separating what actually mattered from the noise.

But the most interesting discovery wasn’t the output.

It was tone.

When the tone was wrong, too robotic, too clever, too interrogative, it actually made things worse. It felt like being questioned or judged.

But when the tone was calm and human something different happened.

You slowed down. You explained things better. You felt guided. And when that happened, clarity started to appear.

That moment stuck with me.

Because at the same time I kept noticing the exact same pattern everywhere else. In Facebook groups, community forums, and conversations with friends. Someone would write a long post about a dispute with a landlord, an employer, a neighbour, or a contract. They were clearly trying to explain what had happened, but the explanation was messy. Important facts halfway down the post. Dates missing. Emotion mixed with facts.

Then the replies would start arriving. Opinions, arguments, conflicting advice. Very rarely did the situation actually become clearer. Watching that again and again made something obvious.

Most people don’t struggle with legal problems alone. They struggle explaining them. And if you can’t explain your situation clearly, the next step becomes much harder.

That idea slowly turned into BriefBase.

But once the idea existed the real challenge started. I knew something important. If BriefBase was going to work it needed to feel human. Not pretending to be human, that creates trust problems immediately, but supportive, calm and grounded.

Like a friend you talk to. Someone you lean on. That arm-around-your-shoulder feeling that helps you think clearly rather than interrogates you.

Very early in the process I was lucky through a random moment that put Mark Thelwall at 1UPBrand and myself together. We clicked quickly and he understood instantly what I was trying to build and how I envisioned BriefBase. Without Mark's support, experience, and dedication to the vision, I'm not sure I would have got this far. One of the first things he advised was focusing on the brand identity early.

That might sound unusual when you’re still at the idea stage, but it helped enormously. Having a strong sense of the brand early grounded everything. It stopped the project drifting into generic “AI tool” territory and kept the focus on clarity, trust and simplicity.

At the same time I was working with an AI developer trying to shape how the conversation actually worked. And this turned out to be much harder than expected.

Getting AI to produce answers is easy. Getting it to guide someone through a legal situation without sounding robotic, overly legal, or like it’s pretending to be human is incredibly difficult. That’s not even mentioning the guardrails, GDPR, disclaimers and backend architecture required just to get an MVP launched. That part is a whole story on its own.

Back to the conversation itself.

If it sounded too mechanical people disengaged. Too empathetic and it becomes irritating. Not enough and the robot gives itself away.

There is a very narrow line where something feels supportive without pretending to be something it isn’t.

Finding that line eventually led me to Lena Shakurova founder of ParsLabs and a leading figure in AI development. What followed was about four months of intense work led by Lena, with the three of us involved. Testing conversations, breaking them, rebuilding them, adjusting tone and adjusting structure. Trying to create something that guides the user rather than letting the user unknowingly guide the AI down the wrong path.

Because legal conversations are different. They need structure, the right questions at the right time, and clarity to emerge gradually. Alongside this another important part of BriefBase started to take shape.

The brief itself.

Because if someone is going to explain their situation clearly the output has to make sense to a solicitor reading it.

That part wasn’t built in isolation. Input from legal professionals helped shape how the brief should be structured so that it reads in a way that is familiar and useful to someone in practice. People like Safi were hugely helpful in this stage, helping test the structure and making sure the end result wasn’t just organised information but something that actually works in a legal context.

And then one day something changed.

The interaction finally felt right.

The BriefBase persona appeared. Not a character, and not a chatbot. I don’t see BriefBase that way and never wanted it positioned like that. But a calm guiding presence that helps someone explain their situation without rushing them or overwhelming them.

That moment was huge for all of us because suddenly it worked.

You could explain something messy and feel the conversation gradually bringing clarity to the situation. That was the moment BriefBase really started to exist. Not just as an idea, but as something real.

What exists today is version one. An MVP.

There’s still a huge amount to improve and build, but reaching this point has already been a rewarding journey. It all started with a legal problem I struggled to explain, a light bulb moment about tone in AI, and many long hours building the idea from my bunk offshore. That eventually became BriefBase.

If you’re curious about the project we’ve been building, you can see it here:

https://briefbase.ai (currently testing with UK legal processes)

Has anyone here ever struggled to explain a situation clearly to a solicitor, or worried you might miss something important?