From Chat to Deliverable to Debate: The 3 Ways to Work With an AI Expert

Most people's experience of AI assistance is a single chat window. You type a question, you get an answer, you refine it. That's one legitimate way to work with an AI expert, but it's not the only one, and treating it as the only one means missing tools that fit other kinds of problems better. There are three distinct ways to work with an AI expert on Cultivaition, and each is suited to a different shape of problem.

1:1 Chat Mode

Chat is the lowest-friction mode, and it costs the least: 1 token per conversation. It's the right choice when you have a question, not a project — something you want to think through out loud, get a quick opinion on, or clarify before deciding whether it's worth turning into something bigger.

Because each of Cultivaition's 500+ experts is built from a synthetic CV with an extracted skillset and behavioural rules, chatting with a specific expert gets you a more focused answer than a generic AI query would. Asking a security-focused expert whether your authentication approach has an obvious flaw is a narrower, more useful question than asking a general-purpose assistant to "review my app," because the expert's behavioural rules shape it to focus on that lane specifically.

Chat also benefits from persona memory: each expert can remember up to 10 notes about you and your project across sessions. That means a chat today can build on a chat from last week without you re-explaining context, which starts to make the mode feel less like a one-off query tool and more like an ongoing working relationship with a specific expert.

Chat is not the right mode when you need a finished artifact to hand to someone else, or when you specifically want to see disagreement between different perspectives. For those, the other two modes exist.

Deliverable Mode

Deliverable mode exists for when the output itself matters, not just the conversation that produces it. Cultivaition supports exactly three deliverable types: a website audit, a PRD draft, and an architecture diagram. Each deliverable costs 8 tokens, reflecting the greater depth and structure involved compared to a chat exchange.

The distinguishing feature of deliverable mode is that the output is a self-contained artifact, not a conversation transcript. A PRD draft is something you can hand to a developer or a design partner directly. An architecture diagram is something you can put in front of a technical hire as a starting point for discussion. A website audit is something you can work through as a checklist. Deliverables can also be shared via a public shareable URL, which matters in practice — it means you can send a co-founder or an investor a link rather than exporting a file and emailing it around.

Deliverable mode is the right choice when you already know roughly what you need and want it structured properly the first time, rather than assembled piece by piece through back-and-forth chat. It is the wrong choice when you're still exploring a question and don't yet know what shape the answer should take — that's better handled in chat first, then formalized into a deliverable once you know what you're asking for.

It's worth repeating here what applies to every mode: a deliverable from Cultivaition is a draft to review, not a guaranteed-correct final artifact. None of the three deliverable types come with a claim of human-expert-level quality assurance. Treat them as a strong starting point, not a finished product.

Society Room Mode

Society Room mode is the most distinct of the three, because it introduces something chat and deliverables don't: multiple perspectives in the same session. A Society Room brings together 2 to 6 experts to discuss a topic, either through structured moderated rounds or through a free conversation where you approve who speaks next and can @mention specific experts to bring them in directly. Each round costs 4 tokens.

This mode exists for the specific situation where a single answer would flatten a real tradeoff. If you ask one expert whether to prioritize speed or robustness in a launch decision, you get one framing. If you put a growth-minded expert and an engineering-minded expert in a Society Room together, you see where they actually disagree, and why — which is usually more useful than either opinion alone, because it mirrors the kind of tension a real leadership team would surface.

Society Rooms are the right mode when a decision genuinely involves competing priorities across disciplines, and you want to see that tension made explicit rather than pre-resolved by a single response. They're the wrong mode for a simple factual question, where the cost and complexity of a multi-expert discussion isn't warranted — chat handles that better and more cheaply.

Which Mode Fits Which Problem

A useful way to decide is to ask what you actually have in hand before you start. If you have a question, use chat. If you have a clear need for one of the three specific artifacts — a website audit, a PRD draft, or an architecture diagram — use deliverable mode. If you have a decision that involves real tradeoffs across more than one discipline and want to see the disagreement laid out, use a Society Room.

These modes aren't mutually exclusive in a single project. A common pattern is to start in chat to clarify a rough idea, escalate to a deliverable once the idea is concrete enough to formalize, and bring in a Society Room specifically for the one or two decision points in that deliverable where reasonable experts would actually disagree. Used together this way, the three modes cover most of the ground a founder needs from an AI expert — while remaining, at every step, a set of drafts and discussions for you to review and decide on, not a substitute for your own judgment.