Write Studio: Turning Client Briefs into Proposals Faster.

The proposal bottleneck every design team knows

Briefs never arrive in the shape you need them. A client call gets recorded and half-transcribed. A moodboard link lands in Slack with no context. Three bullet points appear in an email subject line: "budget", "timeline??", "can we see something by Friday." Somewhere between that and a document a client will actually sign sits a stretch of work that nobody enjoys: re-reading the mess, structuring it, and writing scope, deliverables, timeline and investment sections from a blank page.

Most studios already lean on AI somewhere in that process. It's common practice now to drop a call transcript into a general chat window, get a rough summary back, then copy-paste that into a separate document to polish. That workflow works, but it fragments across three or four tabs: a notes app for the raw brief, a chat window for drafting, a document for formatting, and often a second chat window when the first model's tone falls flat. Every switch is a chance to lose context, paste the wrong version, or forget which draft had the client's actual budget figure in it.

The fix isn't a smarter template. It's one surface where a messy brief can be summarised, restructured and refined without the file ever leaving the page.

Inside Write: what the model picker actually gives you

Write is Stensyl's long-form drafting surface, and its model picker carries six writing models on every plan, no tier gating: GPT-5.6 Luna, GPT-5.6 Terra and GPT-5.6 Sol from OpenAI, Gemini Flash and Gemini Pro from Google, and Claude Sonnet 5 and Claude Opus 4.8 from Anthropic. All eight, including Claude Fable 5, are available in the Canvas Creative Assistant node for teams building proposal generation into a repeatable pipeline.

That range matters because a proposal isn't one writing task, it's several stacked on top of each other. Turning a scrappy brief into a structured outline is a fast, low-reasoning job: a model like Luna or Gemini Flash can chew through a call transcript and spit out goals, constraints and deliverables in seconds. Writing the section that justifies your pricing logic, or flags a scope risk the client hasn't considered, is a different job entirely. That's where GPT-5.6 Sol or Claude Opus 4.8 earn their place: more deliberate reasoning, better at holding nuance across a longer section, better at catching the gap a faster model glossed over.

The part that actually changes the workflow is that the picker sits inside the same document. There's no exporting a summary out of one tool and pasting it into another. You draft the outline with Luna, then switch to Opus 4.8 for the scope section, then back to Sonnet 5 for a plain-language pass, all inside the one Write document, all without retyping the brief each time.

A repeatable workflow: brief to proposal in four passes

The most useful way to think about Write for proposals isn't "one prompt, one document." It's four distinct passes, each suited to a different kind of model work.

Pass 1: Structured summary

Paste the raw brief in whatever state it exists, call notes, an email thread, a slide excerpt from the client's brand deck, and ask a fast model (Luna or Gemini Flash) to produce a structured summary: goals, constraints, deliverables, anything the client mentioned about budget or timing. This is the pass that turns "messy" into "readable" without losing information.

Pass 2: Proposal skeleton

From that summary, generate the proposal skeleton itself: scope, timeline, deliverables, investment. This is standard proposal architecture, and it holds regardless of discipline, because a client signing off on a kitchen renovation and a client signing off on a 15-second brand sting both need to see exactly what they're getting, when, and for how much.

Pass 3: Precision pass

Switch models. A reasoning-heavy model like GPT-5.6 Sol or Claude Opus 4.8 rereads the skeleton and tightens the scope language, specifically hunting for the gaps a faster model tends to skip: unstated assumptions, ambiguous deliverable counts, revision limits that were never mentioned. This is the pass that turns a passable draft into one that protects the studio if scope creeps later.

Pass 4: Brand sanity check

Route the final draft through Ray, Stensyl's assistant, for a check against the project's brand identity and prior client history stored in Projects. Ray can flag tone mismatches, inconsistent terminology, or a proposal that reads nothing like the last one this client received.

Discipline examples: the same workflow, different proposals

The four-pass structure holds across very different creative disciplines. What changes is the content of the skeleton, not the shape of the workflow.

• Interior design: A client walkthrough transcript, full of half-finished sentences about a kitchen they hate and a hallway they love, becomes a room-by-room scope. Pass 1 pulls out every space mentioned and what the client said about it. Pass 2 turns that into sections with material notes and phased timelines per room. Pass 3 checks that the material language is specific enough to price against.
• Motion design: A marketing team's one-line brief, "need a 15-second brand sting, punchy, for socials", expands into a full production proposal: shot count, revision rounds, delivery formats for different platforms. The fast pass structures the ask; the precision pass adds the caveats a client rarely thinks to ask about, like how many revision rounds are included before extra costs kick in.
• Marketing and advertising: An ad brief becomes a campaign proposal with clearly separated sections for social carousels and paid formats, the kind of split that later feeds straight into Marketing Studio when it's time to actually build the creative. Keeping these sections distinct in the proposal avoids the common confusion where a client assumes one deliverable covers both organic and paid.
• Exhibition design: A sparse stand brief, just a footprint number and a footfall target, expands into a full concept proposal with build phases and budget bands. This is the discipline where the fast-to-reasoning switch matters most: Pass 1 structures the given numbers, Pass 3 is where a reasoning model works out what those numbers imply for build complexity and flags where the budget band might be unrealistic.

Where Write fits with the rest of the proposal pipeline

A written proposal rarely travels alone. Most clients expect at least a hint of what the finished work will look like, and Write is built to sit inside a pipeline rather than stand apart from it.

Supporting visuals can be pulled from Image or Boards without leaving the project: a moodboard collected in Boards for an interior scheme, a first-frame render generated in Image for a product concept, a reference collage for an exhibition stand. Because Boards now merges the old moodboard and storyboard functions into one canvas, the same surface that held early references can also hold the start/end frames for any video content the proposal promises.

Before drafting even starts, Research, Stensyl's Perplexity-backed surface, can ground the proposal in real market or competitor context: pricing benchmarks, category trends, a competitor's recent campaign. That keeps claims in the proposal defensible rather than guessed.

All of this stays inside a shared Project, so the brief, the drafts, and the final proposal live in one place for the whole team rather than scattered across someone's downloads folder and a Slack thread. And Ray, sitting in the same project chat, can act as a second reader at any point, not just at the final pass, checking language against the brand identity the team has already set up for that client.

Pass	Model type	Job
1. Summary	Fast (Luna, Gemini Flash)	Structure raw brief into goals, constraints, deliverables
2. Skeleton	Fast to balanced (Terra, Sonnet 5)	Build scope, timeline, deliverables, investment sections
3. Precision	Reasoning-heavy (Sol, Opus 4.8)	Tighten scope, flag gaps, sharpen pricing rationale
4. Sanity check	Ray	Check tone and brand consistency against Project history

The takeaway

A client brief doesn't need five tools to become a proposal. It needs one surface with the right model for each pass: something fast to handle structure, something with sharper reasoning to handle nuance, and a way to move between them without breaking flow or retyping the brief for the third time. Write is built around that switch, six models on every plan, sitting inside one document rather than scattered across separate chat tabs.

The result isn't just a faster first draft. It's fewer version-control headaches across the team, fewer moments where someone asks "wait, which draft has the updated timeline?", and proposals that go out while the brief is still fresh rather than a week later once someone's finally found time to stitch it all together.