Why We Killed the Standalone Chat

For most of this platform's life, a chat in OBEL worked the way a chat works everywhere else: you opened a window, you talked to a model, and what happened in that window was yours. Nobody else could see it. Nothing about it was organised relative to anything else. When you moved on, so did the context, the reasoning, and whatever the model had figured out along the way.

That is a familiar shape for a consumer chatbot. It is a real liability for an organisation trying to govern how AI gets used. We removed it. Every conversation in OBEL now belongs to a Space, and the standalone chat is gone.

The knowledge that died with the tab

A standalone conversation has exactly one owner, and it stops existing, for practical purposes, the moment that person stops looking at it. Nobody else on the team can pick up where it left off. Nobody can audit it as part of a broader pattern, because it was never part of anything broader to begin with. Five people can ask an AI the same underlying question in five separate windows, get five slightly different answers, and there will be no record that any of them were even related.

That is not a UX inconvenience. It is an audit trail with holes deliberately built into it, one per person, by design. An organisation that cannot say what its people asked an AI, in relation to what project, reviewed by whom, is not running a governed AI program. It is running a very large number of ungoverned ones that happen to share a login page.

A Space is a governed container, not a folder

The obvious fix looks like adding folders: let people organise their own chats into groups. We did not build that, because a folder does not change who owns the conversation or who can see it. It just tidies up the mess without touching the actual problem.

A Space is a different kind of object. Every conversation inside it is visible to every member with access, under one permission model, with one shared audit trail. A Space can carry its own memory, its own default persona, its own attached Knowledge Bases, scoped to exactly the people who should see them. Nothing about that is a display preference. It is who is allowed to see what, enforced at the point conversations are created, not reconstructed afterward from scattered chat logs.

Spaces carry policy, not just organisation

The clearest example of what this unlocks is something we shipped alongside it: Response Mode, OBEL's multi-model comparison feature, is no longer a private toggle a user flips per message. It is a property of the Space.

Multi-model response runs a primary model and a review set in parallel on every turn, compares the results, and returns one answer with the comparison trail attached. It is materially more thorough, and it costs more and takes longer per message. Leaving that decision entirely to individual discretion, message by message, is exactly the kind of silent policy erosion we do not accept anywhere else in this platform. Nobody gets to quietly turn off PII scrubbing because it is inconvenient for one message. Response Mode should not be different just because the stakes look lower.

So it lives on the Space. A legal review Space can default to multi-model comparison for every conversation inside it. A general working Space can default to single-model and stay fast. Escalating a Space's default to multi-model either applies immediately, for a member the org has already trusted with that entitlement, or it files a request that an org admin has to actually approve, surfaced directly in their notification centre alongside every other decision that needs a human. De-escalating back to single is always free, because reducing assurance never needs a gatekeeper. Only raising it does.

That is the pattern we want more of. A Space is not just a place where conversations happen to be grouped. It is where the policy that governs those conversations actually lives, inherited by everything created inside it, changeable only by whoever is supposed to be allowed to change it.

“A conversation nobody else can see, governed by nobody but the person who opened it, was never really governed at all. It was just private. Spaces are what it looks like when we stopped pretending those were the same thing.”

- Denis Bouton

Denis Bouton is the founding Chief Architect of OBEL™ and Managing Partner at ninthLABS Ventures, where he advises Post-Sales Services and Customer Success organisations on scaling onboarding, adoption, and retention. He leads OBEL's product and platform strategy.